ViewVC Help
View File | Revision Log | Show Annotations | Download File | Root Listing
root/jsr166/jsr166/src/jsr166y/ForkJoinPool.java
(Generate patch)

Comparing jsr166/src/jsr166y/ForkJoinPool.java (file contents):
Revision 1.16 by jsr166, Thu Jul 23 19:44:46 2009 UTC vs.
Revision 1.63 by dl, Fri Aug 13 16:21:23 2010 UTC

# Line 5 | Line 5
5   */
6  
7   package jsr166y;
8 < import java.util.*;
8 >
9   import java.util.concurrent.*;
10 < import java.util.concurrent.locks.*;
11 < import java.util.concurrent.atomic.*;
12 < import sun.misc.Unsafe;
13 < import java.lang.reflect.*;
10 >
11 > import java.util.ArrayList;
12 > import java.util.Arrays;
13 > import java.util.Collection;
14 > import java.util.Collections;
15 > import java.util.List;
16 > import java.util.concurrent.locks.LockSupport;
17 > import java.util.concurrent.locks.ReentrantLock;
18 > import java.util.concurrent.atomic.AtomicInteger;
19 > import java.util.concurrent.CountDownLatch;
20  
21   /**
22 < * An {@link ExecutorService} for running {@link ForkJoinTask}s.  A
23 < * ForkJoinPool provides the entry point for submissions from
24 < * non-ForkJoinTasks, as well as management and monitoring operations.
25 < * Normally a single ForkJoinPool is used for a large number of
20 < * submitted tasks. Otherwise, use would not usually outweigh the
21 < * construction and bookkeeping overhead of creating a large set of
22 < * threads.
22 > * An {@link ExecutorService} for running {@link ForkJoinTask}s.
23 > * A {@code ForkJoinPool} provides the entry point for submissions
24 > * from non-{@code ForkJoinTask} clients, as well as management and
25 > * monitoring operations.
26   *
27 < * <p>ForkJoinPools differ from other kinds of Executors mainly in
28 < * that they provide <em>work-stealing</em>: all threads in the pool
29 < * attempt to find and execute subtasks created by other active tasks
30 < * (eventually blocking if none exist). This makes them efficient when
31 < * most tasks spawn other subtasks (as do most ForkJoinTasks), as well
32 < * as the mixed execution of some plain Runnable- or Callable- based
33 < * activities along with ForkJoinTasks. When setting
34 < * {@code setAsyncMode}, a ForkJoinPools may also be appropriate for
35 < * use with fine-grained tasks that are never joined. Otherwise, other
33 < * ExecutorService implementations are typically more appropriate
34 < * choices.
27 > * <p>A {@code ForkJoinPool} differs from other kinds of {@link
28 > * ExecutorService} mainly by virtue of employing
29 > * <em>work-stealing</em>: all threads in the pool attempt to find and
30 > * execute subtasks created by other active tasks (eventually blocking
31 > * waiting for work if none exist). This enables efficient processing
32 > * when most tasks spawn other subtasks (as do most {@code
33 > * ForkJoinTask}s). When setting <em>asyncMode</em> to true in
34 > * constructors, {@code ForkJoinPool}s may also be appropriate for use
35 > * with event-style tasks that are never joined.
36   *
37 < * <p>A ForkJoinPool may be constructed with a given parallelism level
38 < * (target pool size), which it attempts to maintain by dynamically
39 < * adding, suspending, or resuming threads, even if some tasks are
40 < * waiting to join others. However, no such adjustments are performed
41 < * in the face of blocked IO or other unmanaged synchronization. The
42 < * nested {@code ManagedBlocker} interface enables extension of
43 < * the kinds of synchronization accommodated.  The target parallelism
44 < * level may also be changed dynamically ({@code setParallelism})
45 < * and thread construction can be limited using methods
45 < * {@code setMaximumPoolSize} and/or
46 < * {@code setMaintainsParallelism}.
37 > * <p>A {@code ForkJoinPool} is constructed with a given target
38 > * parallelism level; by default, equal to the number of available
39 > * processors. The pool attempts to maintain enough active (or
40 > * available) threads by dynamically adding, suspending, or resuming
41 > * internal worker threads, even if some tasks are stalled waiting to
42 > * join others. However, no such adjustments are guaranteed in the
43 > * face of blocked IO or other unmanaged synchronization. The nested
44 > * {@link ManagedBlocker} interface enables extension of the kinds of
45 > * synchronization accommodated.
46   *
47   * <p>In addition to execution and lifecycle control methods, this
48   * class provides status check methods (for example
49 < * {@code getStealCount}) that are intended to aid in developing,
49 > * {@link #getStealCount}) that are intended to aid in developing,
50   * tuning, and monitoring fork/join applications. Also, method
51 < * {@code toString} returns indications of pool state in a
51 > * {@link #toString} returns indications of pool state in a
52   * convenient form for informal monitoring.
53   *
54 + * <p> As is the case with other ExecutorServices, there are three
55 + * main task execution methods summarized in the following
56 + * table. These are designed to be used by clients not already engaged
57 + * in fork/join computations in the current pool.  The main forms of
58 + * these methods accept instances of {@code ForkJoinTask}, but
59 + * overloaded forms also allow mixed execution of plain {@code
60 + * Runnable}- or {@code Callable}- based activities as well.  However,
61 + * tasks that are already executing in a pool should normally
62 + * <em>NOT</em> use these pool execution methods, but instead use the
63 + * within-computation forms listed in the table.
64 + *
65 + * <table BORDER CELLPADDING=3 CELLSPACING=1>
66 + *  <tr>
67 + *    <td></td>
68 + *    <td ALIGN=CENTER> <b>Call from non-fork/join clients</b></td>
69 + *    <td ALIGN=CENTER> <b>Call from within fork/join computations</b></td>
70 + *  </tr>
71 + *  <tr>
72 + *    <td> <b>Arange async execution</td>
73 + *    <td> {@link #execute(ForkJoinTask)}</td>
74 + *    <td> {@link ForkJoinTask#fork}</td>
75 + *  </tr>
76 + *  <tr>
77 + *    <td> <b>Await and obtain result</td>
78 + *    <td> {@link #invoke(ForkJoinTask)}</td>
79 + *    <td> {@link ForkJoinTask#invoke}</td>
80 + *  </tr>
81 + *  <tr>
82 + *    <td> <b>Arrange exec and obtain Future</td>
83 + *    <td> {@link #submit(ForkJoinTask)}</td>
84 + *    <td> {@link ForkJoinTask#fork} (ForkJoinTasks <em>are</em> Futures)</td>
85 + *  </tr>
86 + * </table>
87 + *
88 + * <p><b>Sample Usage.</b> Normally a single {@code ForkJoinPool} is
89 + * used for all parallel task execution in a program or subsystem.
90 + * Otherwise, use would not usually outweigh the construction and
91 + * bookkeeping overhead of creating a large set of threads. For
92 + * example, a common pool could be used for the {@code SortTasks}
93 + * illustrated in {@link RecursiveAction}. Because {@code
94 + * ForkJoinPool} uses threads in {@linkplain java.lang.Thread#isDaemon
95 + * daemon} mode, there is typically no need to explicitly {@link
96 + * #shutdown} such a pool upon program exit.
97 + *
98 + * <pre>
99 + * static final ForkJoinPool mainPool = new ForkJoinPool();
100 + * ...
101 + * public void sort(long[] array) {
102 + *   mainPool.invoke(new SortTask(array, 0, array.length));
103 + * }
104 + * </pre>
105 + *
106   * <p><b>Implementation notes</b>: This implementation restricts the
107   * maximum number of running threads to 32767. Attempts to create
108 < * pools with greater than the maximum result in
109 < * IllegalArgumentExceptions.
108 > * pools with greater than the maximum number result in
109 > * {@code IllegalArgumentException}.
110 > *
111 > * <p>This implementation rejects submitted tasks (that is, by throwing
112 > * {@link RejectedExecutionException}) only when the pool is shut down
113 > * or internal resources have been exhausted.
114   *
115   * @since 1.7
116   * @author Doug Lea
# Line 63 | Line 118 | import java.lang.reflect.*;
118   public class ForkJoinPool extends AbstractExecutorService {
119  
120      /*
121 <     * See the extended comments interspersed below for design,
122 <     * rationale, and walkthroughs.
123 <     */
124 <
125 <    /** Mask for packing and unpacking shorts */
126 <    private static final int  shortMask = 0xffff;
127 <
128 <    /** Max pool size -- must be a power of two minus 1 */
129 <    private static final int MAX_THREADS =  0x7FFF;
130 <
131 <    /**
132 <     * Factory for creating new ForkJoinWorkerThreads.  A
133 <     * ForkJoinWorkerThreadFactory must be defined and used for
134 <     * ForkJoinWorkerThread subclasses that extend base functionality
135 <     * or initialize threads with different contexts.
121 >     * Implementation Overview
122 >     *
123 >     * This class provides the central bookkeeping and control for a
124 >     * set of worker threads: Submissions from non-FJ threads enter
125 >     * into a submission queue. Workers take these tasks and typically
126 >     * split them into subtasks that may be stolen by other workers.
127 >     * The main work-stealing mechanics implemented in class
128 >     * ForkJoinWorkerThread give first priority to processing tasks
129 >     * from their own queues (LIFO or FIFO, depending on mode), then
130 >     * to randomized FIFO steals of tasks in other worker queues, and
131 >     * lastly to new submissions. These mechanics do not consider
132 >     * affinities, loads, cache localities, etc, so rarely provide the
133 >     * best possible performance on a given machine, but portably
134 >     * provide good throughput by averaging over these factors.
135 >     * (Further, even if we did try to use such information, we do not
136 >     * usually have a basis for exploiting it. For example, some sets
137 >     * of tasks profit from cache affinities, but others are harmed by
138 >     * cache pollution effects.)
139 >     *
140 >     * Beyond work-stealing support and essential bookkeeping, the
141 >     * main responsibility of this framework is to take actions when
142 >     * one worker is waiting to join a task stolen (or always held by)
143 >     * another.  Becauae we are multiplexing many tasks on to a pool
144 >     * of workers, we can't just let them block (as in Thread.join).
145 >     * We also cannot just reassign the joiner's run-time stack with
146 >     * another and replace it later, which would be a form of
147 >     * "continuation", that even if possible is not necessarily a good
148 >     * idea. Given that the creation costs of most threads on most
149 >     * systems mainly surrounds setting up runtime stacks, thread
150 >     * creation and switching is usually not much more expensive than
151 >     * stack creation and switching, and is more flexible). Instead we
152 >     * combine two tactics:
153 >     *
154 >     *   Helping: Arranging for the joiner to execute some task that it
155 >     *      would be running if the steal had not occurred.  Method
156 >     *      ForkJoinWorkerThread.helpJoinTask tracks joining->stealing
157 >     *      links to try to find such a task.
158 >     *
159 >     *   Compensating: Unless there are already enough live threads,
160 >     *      method helpMaintainParallelism() may create or or
161 >     *      re-activate a spare thread to compensate for blocked
162 >     *      joiners until they unblock.
163 >     *
164 >     * Because the determining existence of conservatively safe
165 >     * helping targets, the availability of already-created spares,
166 >     * and the apparent need to create new spares are all racy and
167 >     * require heuristic guidance, we rely on multiple retries of
168 >     * each. Further, because it is impossible to keep exactly the
169 >     * target (parallelism) number of threads running at any given
170 >     * time, we allow compensation during joins to fail, and enlist
171 >     * all other threads to help out whenever they are not otherwise
172 >     * occupied (i.e., mainly in method preStep).
173 >     *
174 >     * The ManagedBlocker extension API can't use helping so relies
175 >     * only on compensation in method awaitBlocker.
176 >     *
177 >     * The main throughput advantages of work-stealing stem from
178 >     * decentralized control -- workers mostly steal tasks from each
179 >     * other. We do not want to negate this by creating bottlenecks
180 >     * implementing other management responsibilities. So we use a
181 >     * collection of techniques that avoid, reduce, or cope well with
182 >     * contention. These entail several instances of bit-packing into
183 >     * CASable fields to maintain only the minimally required
184 >     * atomicity. To enable such packing, we restrict maximum
185 >     * parallelism to (1<<15)-1 (enabling twice this (to accommodate
186 >     * unbalanced increments and decrements) to fit into a 16 bit
187 >     * field, which is far in excess of normal operating range.  Even
188 >     * though updates to some of these bookkeeping fields do sometimes
189 >     * contend with each other, they don't normally cache-contend with
190 >     * updates to others enough to warrant memory padding or
191 >     * isolation. So they are all held as fields of ForkJoinPool
192 >     * objects.  The main capabilities are as follows:
193 >     *
194 >     * 1. Creating and removing workers. Workers are recorded in the
195 >     * "workers" array. This is an array as opposed to some other data
196 >     * structure to support index-based random steals by workers.
197 >     * Updates to the array recording new workers and unrecording
198 >     * terminated ones are protected from each other by a lock
199 >     * (workerLock) but the array is otherwise concurrently readable,
200 >     * and accessed directly by workers. To simplify index-based
201 >     * operations, the array size is always a power of two, and all
202 >     * readers must tolerate null slots. Currently, all worker thread
203 >     * creation is on-demand, triggered by task submissions,
204 >     * replacement of terminated workers, and/or compensation for
205 >     * blocked workers. However, all other support code is set up to
206 >     * work with other policies.
207 >     *
208 >     * To ensure that we do not hold on to worker references that
209 >     * would prevent GC, ALL accesses to workers are via indices into
210 >     * the workers array (which is one source of some of the unusual
211 >     * code constructions here). In essence, the workers array serves
212 >     * as a WeakReference mechanism. Thus for example the event queue
213 >     * stores worker indices, not worker references. Access to the
214 >     * workers in associated methods (for example releaseEventWaiters)
215 >     * must both index-check and null-check the IDs. All such accesses
216 >     * ignore bad IDs by returning out early from what they are doing,
217 >     * since this can only be associated with shutdown, in which case
218 >     * it is OK to give up. On termination, we just clobber these
219 >     * data structures without trying to use them.
220 >     *
221 >     * 2. Bookkeeping for dynamically adding and removing workers. We
222 >     * aim to approximately maintain the given level of parallelism.
223 >     * When some workers are known to be blocked (on joins or via
224 >     * ManagedBlocker), we may create or resume others to take their
225 >     * place until they unblock (see below). Implementing this
226 >     * requires counts of the number of "running" threads (i.e., those
227 >     * that are neither blocked nor artifically suspended) as well as
228 >     * the total number.  These two values are packed into one field,
229 >     * "workerCounts" because we need accurate snapshots when deciding
230 >     * to create, resume or suspend.  Note however that the
231 >     * correspondance of these counts to reality is not guaranteed. In
232 >     * particular updates for unblocked threads may lag until they
233 >     * actually wake up.
234 >     *
235 >     * 3. Maintaining global run state. The run state of the pool
236 >     * consists of a runLevel (SHUTDOWN, TERMINATING, etc) similar to
237 >     * those in other Executor implementations, as well as a count of
238 >     * "active" workers -- those that are, or soon will be, or
239 >     * recently were executing tasks. The runLevel and active count
240 >     * are packed together in order to correctly trigger shutdown and
241 >     * termination. Without care, active counts can be subject to very
242 >     * high contention.  We substantially reduce this contention by
243 >     * relaxing update rules.  A worker must claim active status
244 >     * prospectively, by activating if it sees that a submitted or
245 >     * stealable task exists (it may find after activating that the
246 >     * task no longer exists). It stays active while processing this
247 >     * task (if it exists) and any other local subtasks it produces,
248 >     * until it cannot find any other tasks. It then tries
249 >     * inactivating (see method preStep), but upon update contention
250 >     * instead scans for more tasks, later retrying inactivation if it
251 >     * doesn't find any.
252 >     *
253 >     * 4. Managing idle workers waiting for tasks. We cannot let
254 >     * workers spin indefinitely scanning for tasks when none are
255 >     * available. On the other hand, we must quickly prod them into
256 >     * action when new tasks are submitted or generated.  We
257 >     * park/unpark these idle workers using an event-count scheme.
258 >     * Field eventCount is incremented upon events that may enable
259 >     * workers that previously could not find a task to now find one:
260 >     * Submission of a new task to the pool, or another worker pushing
261 >     * a task onto a previously empty queue.  (We also use this
262 >     * mechanism for termination actions that require wakeups of idle
263 >     * workers).  Each worker maintains its last known event count,
264 >     * and blocks when a scan for work did not find a task AND its
265 >     * lastEventCount matches the current eventCount. Waiting idle
266 >     * workers are recorded in a variant of Treiber stack headed by
267 >     * field eventWaiters which, when nonzero, encodes the thread
268 >     * index and count awaited for by the worker thread most recently
269 >     * calling eventSync. This thread in turn has a record (field
270 >     * nextEventWaiter) for the next waiting worker.  In addition to
271 >     * allowing simpler decisions about need for wakeup, the event
272 >     * count bits in eventWaiters serve the role of tags to avoid ABA
273 >     * errors in Treiber stacks.  To reduce delays in task diffusion,
274 >     * workers not otherwise occupied may invoke method
275 >     * releaseEventWaiters, that removes and signals (unparks) workers
276 >     * not waiting on current count. To reduce stalls, To minimize
277 >     * task production stalls associate with signalling, any worker
278 >     * pushing a task on an empty queue invokes the weaker method
279 >     * signalWork, that only releases idle workers until it detects
280 >     * interference by other threads trying to release, and lets them
281 >     * take over.  The net effect is a tree-like diffusion of signals,
282 >     * where released threads (and possibly others) help with unparks.
283 >     * To further reduce contention effects a bit, failed CASes to
284 >     * increment field eventCount are tolerated without retries.
285 >     * Conceptually they are merged into the same event, which is OK
286 >     * when their only purpose is to enable workers to scan for work.
287 >     *
288 >     * 5. Managing suspension of extra workers. When a worker is about
289 >     * to block waiting for a join (or via ManagedBlockers), we may
290 >     * create a new thread to maintain parallelism level, or at least
291 >     * avoid starvation. Usually, extra threads are needed for only
292 >     * very short periods, yet join dependencies are such that we
293 >     * sometimes need them in bursts. Rather than create new threads
294 >     * each time this happens, we suspend no-longer-needed extra ones
295 >     * as "spares". For most purposes, we don't distinguish "extra"
296 >     * spare threads from normal "core" threads: On each call to
297 >     * preStep (the only point at which we can do this) a worker
298 >     * checks to see if there are now too many running workers, and if
299 >     * so, suspends itself.  Method helpMaintainParallelism looks for
300 >     * suspended threads to resume before considering creating a new
301 >     * replacement. The spares themselves are encoded on another
302 >     * variant of a Treiber Stack, headed at field "spareWaiters".
303 >     * Note that the use of spares is intrinsically racy.  One thread
304 >     * may become a spare at about the same time as another is
305 >     * needlessly being created. We counteract this and related slop
306 >     * in part by requiring resumed spares to immediately recheck (in
307 >     * preStep) to see whether they they should re-suspend.  To avoid
308 >     * long-term build-up of spares, the oldest spare (see
309 >     * ForkJoinWorkerThread.suspendAsSpare) occasionally wakes up if
310 >     * not signalled and calls tryTrimSpare, which uses two different
311 >     * thresholds: Always killing if the number of spares is greater
312 >     * that 25% of total, and killing others only at a slower rate
313 >     * (UNUSED_SPARE_TRIM_RATE_NANOS).
314 >     *
315 >     * 6. Deciding when to create new workers. The main dynamic
316 >     * control in this class is deciding when to create extra threads
317 >     * in method helpMaintainParallelism. We would like to keep
318 >     * exactly #parallelism threads running, which is an impossble
319 >     * task. We always need to create one when the number of running
320 >     * threads would become zero and all workers are busy. Beyond
321 >     * this, we must rely on heuristics that work well in the the
322 >     * presence of transients phenomena such as GC stalls, dynamic
323 >     * compilation, and wake-up lags. These transients are extremely
324 >     * common -- we are normally trying to fully saturate the CPUs on
325 >     * a machine, so almost any activity other than running tasks
326 >     * impedes accuracy. Our main defense is to allow some slack in
327 >     * creation thresholds, using rules that reflect the fact that the
328 >     * more threads we have running, the more likely that we are
329 >     * underestimating the number running threads. The rules also
330 >     * better cope with the fact that some of the methods in this
331 >     * class tend to never become compiled (but are interpreted), so
332 >     * some components of the entire set of controls might execute 100
333 >     * times faster than others. And similarly for cases where the
334 >     * apparent lack of work is just due to GC stalls and other
335 >     * transient system activity.
336 >     *
337 >     * Beware that there is a lot of representation-level coupling
338 >     * among classes ForkJoinPool, ForkJoinWorkerThread, and
339 >     * ForkJoinTask.  For example, direct access to "workers" array by
340 >     * workers, and direct access to ForkJoinTask.status by both
341 >     * ForkJoinPool and ForkJoinWorkerThread.  There is little point
342 >     * trying to reduce this, since any associated future changes in
343 >     * representations will need to be accompanied by algorithmic
344 >     * changes anyway.
345 >     *
346 >     * Style notes: There are lots of inline assignments (of form
347 >     * "while ((local = field) != 0)") which are usually the simplest
348 >     * way to ensure the required read orderings (which are sometimes
349 >     * critical). Also several occurrences of the unusual "do {}
350 >     * while(!cas...)" which is the simplest way to force an update of
351 >     * a CAS'ed variable. There are also other coding oddities that
352 >     * help some methods perform reasonably even when interpreted (not
353 >     * compiled), at the expense of some messy constructions that
354 >     * reduce byte code counts.
355 >     *
356 >     * The order of declarations in this file is: (1) statics (2)
357 >     * fields (along with constants used when unpacking some of them)
358 >     * (3) internal control methods (4) callbacks and other support
359 >     * for ForkJoinTask and ForkJoinWorkerThread classes, (5) exported
360 >     * methods (plus a few little helpers).
361 >     */
362 >
363 >    /**
364 >     * Factory for creating new {@link ForkJoinWorkerThread}s.
365 >     * A {@code ForkJoinWorkerThreadFactory} must be defined and used
366 >     * for {@code ForkJoinWorkerThread} subclasses that extend base
367 >     * functionality or initialize threads with different contexts.
368       */
369      public static interface ForkJoinWorkerThreadFactory {
370          /**
371           * Returns a new worker thread operating in the given pool.
372           *
373           * @param pool the pool this thread works in
374 <         * @throws NullPointerException if pool is null
374 >         * @throws NullPointerException if the pool is null
375           */
376          public ForkJoinWorkerThread newThread(ForkJoinPool pool);
377      }
378  
379      /**
380 <     * Default ForkJoinWorkerThreadFactory implementation, creates a
380 >     * Default ForkJoinWorkerThreadFactory implementation; creates a
381       * new ForkJoinWorkerThread.
382       */
383 <    static class  DefaultForkJoinWorkerThreadFactory
383 >    static class DefaultForkJoinWorkerThreadFactory
384          implements ForkJoinWorkerThreadFactory {
385          public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
386 <            try {
100 <                return new ForkJoinWorkerThread(pool);
101 <            } catch (OutOfMemoryError oom)  {
102 <                return null;
103 <            }
386 >            return new ForkJoinWorkerThread(pool);
387          }
388      }
389  
# Line 136 | Line 419 | public class ForkJoinPool extends Abstra
419          new AtomicInteger();
420  
421      /**
422 <     * Array holding all worker threads in the pool. Initialized upon
423 <     * first use. Array size must be a power of two.  Updates and
424 <     * replacements are protected by workerLock, but it is always kept
425 <     * in a consistent enough state to be randomly accessed without
426 <     * locking by workers performing work-stealing.
422 >     * Absolute bound for parallelism level. Twice this number plus
423 >     * one (i.e., 0xfff) must fit into a 16bit field to enable
424 >     * word-packing for some counts and indices.
425 >     */
426 >    private static final int MAX_WORKERS   = 0x7fff;
427 >
428 >    /**
429 >     * Array holding all worker threads in the pool.  Array size must
430 >     * be a power of two.  Updates and replacements are protected by
431 >     * workerLock, but the array is always kept in a consistent enough
432 >     * state to be randomly accessed without locking by workers
433 >     * performing work-stealing, as well as other traversal-based
434 >     * methods in this class. All readers must tolerate that some
435 >     * array slots may be null.
436       */
437      volatile ForkJoinWorkerThread[] workers;
438  
439      /**
440 <     * Lock protecting access to workers.
440 >     * Queue for external submissions.
441       */
442 <    private final ReentrantLock workerLock;
442 >    private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
443  
444      /**
445 <     * Condition for awaitTermination.
445 >     * Lock protecting updates to workers array.
446       */
447 <    private final Condition termination;
447 >    private final ReentrantLock workerLock;
448  
449      /**
450 <     * The uncaught exception handler used when any worker
159 <     * abruptly terminates
450 >     * Latch released upon termination.
451       */
452 <    private Thread.UncaughtExceptionHandler ueh;
452 >    private final Phaser termination;
453  
454      /**
455       * Creation factory for worker threads.
# Line 166 | Line 457 | public class ForkJoinPool extends Abstra
457      private final ForkJoinWorkerThreadFactory factory;
458  
459      /**
169     * Head of stack of threads that were created to maintain
170     * parallelism when other threads blocked, but have since
171     * suspended when the parallelism level rose.
172     */
173    private volatile WaitQueueNode spareStack;
174
175    /**
460       * Sum of per-thread steal counts, updated only when threads are
461       * idle or terminating.
462       */
463 <    private final AtomicLong stealCount;
463 >    private volatile long stealCount;
464  
465      /**
466 <     * Queue for external submissions.
466 >     * The last nanoTime that a spare thread was trimmed
467       */
468 <    private final LinkedTransferQueue<ForkJoinTask<?>> submissionQueue;
468 >    private volatile long trimTime;
469  
470      /**
471 <     * Head of Treiber stack for barrier sync. See below for explanation
471 >     * The rate at which to trim unused spares
472       */
473 <    private volatile WaitQueueNode syncStack;
473 >    static final long UNUSED_SPARE_TRIM_RATE_NANOS =
474 >        1000L * 1000L * 1000L; // 1 sec
475  
476      /**
477 <     * The count for event barrier
477 >     * Encoded record of top of treiber stack of threads waiting for
478 >     * events. The top 32 bits contain the count being waited for. The
479 >     * bottom 16 bits contains one plus the pool index of waiting
480 >     * worker thread. (Bits 16-31 are unused.)
481       */
482 <    private volatile long eventCount;
482 >    private volatile long eventWaiters;
483 >
484 >    private static final int  EVENT_COUNT_SHIFT = 32;
485 >    private static final long WAITER_ID_MASK    = (1L << 16) - 1L;
486  
487      /**
488 <     * Pool number, just for assigning useful names to worker threads
488 >     * A counter for events that may wake up worker threads:
489 >     *   - Submission of a new task to the pool
490 >     *   - A worker pushing a task on an empty queue
491 >     *   - termination
492       */
493 <    private final int poolNumber;
493 >    private volatile int eventCount;
494 >
495 >    /**
496 >     * Encoded record of top of treiber stack of spare threads waiting
497 >     * for resumption. The top 16 bits contain an arbitrary count to
498 >     * avoid ABA effects. The bottom 16bits contains one plus the pool
499 >     * index of waiting worker thread.
500 >     */
501 >    private volatile int spareWaiters;
502 >
503 >    private static final int SPARE_COUNT_SHIFT = 16;
504 >    private static final int SPARE_ID_MASK     = (1 << 16) - 1;
505 >
506 >    /**
507 >     * Lifecycle control. The low word contains the number of workers
508 >     * that are (probably) executing tasks. This value is atomically
509 >     * incremented before a worker gets a task to run, and decremented
510 >     * when worker has no tasks and cannot find any.  Bits 16-18
511 >     * contain runLevel value. When all are zero, the pool is
512 >     * running. Level transitions are monotonic (running -> shutdown
513 >     * -> terminating -> terminated) so each transition adds a bit.
514 >     * These are bundled together to ensure consistent read for
515 >     * termination checks (i.e., that runLevel is at least SHUTDOWN
516 >     * and active threads is zero).
517 >     */
518 >    private volatile int runState;
519 >
520 >    // Note: The order among run level values matters.
521 >    private static final int RUNLEVEL_SHIFT     = 16;
522 >    private static final int SHUTDOWN           = 1 << RUNLEVEL_SHIFT;
523 >    private static final int TERMINATING        = 1 << (RUNLEVEL_SHIFT + 1);
524 >    private static final int TERMINATED         = 1 << (RUNLEVEL_SHIFT + 2);
525 >    private static final int ACTIVE_COUNT_MASK  = (1 << RUNLEVEL_SHIFT) - 1;
526 >    private static final int ONE_ACTIVE         = 1; // active update delta
527  
528      /**
529 <     * The maximum allowed pool size
529 >     * Holds number of total (i.e., created and not yet terminated)
530 >     * and running (i.e., not blocked on joins or other managed sync)
531 >     * threads, packed together to ensure consistent snapshot when
532 >     * making decisions about creating and suspending spare
533 >     * threads. Updated only by CAS. Note that adding a new worker
534 >     * requires incrementing both counts, since workers start off in
535 >     * running state.
536       */
537 <    private volatile int maxPoolSize;
537 >    private volatile int workerCounts;
538 >
539 >    private static final int TOTAL_COUNT_SHIFT  = 16;
540 >    private static final int RUNNING_COUNT_MASK = (1 << TOTAL_COUNT_SHIFT) - 1;
541 >    private static final int ONE_RUNNING        = 1;
542 >    private static final int ONE_TOTAL          = 1 << TOTAL_COUNT_SHIFT;
543  
544      /**
545 <     * The desired parallelism level, updated only under workerLock.
545 >     * The target parallelism level.
546 >     * Accessed directly by ForkJoinWorkerThreads.
547       */
548 <    private volatile int parallelism;
548 >    final int parallelism;
549  
550      /**
551       * True if use local fifo, not default lifo, for local polling
552 +     * Read by, and replicated by ForkJoinWorkerThreads
553       */
554 <    private volatile boolean locallyFifo;
554 >    final boolean locallyFifo;
555  
556      /**
557 <     * Holds number of total (i.e., created and not yet terminated)
558 <     * and running (i.e., not blocked on joins or other managed sync)
219 <     * threads, packed into one int to ensure consistent snapshot when
220 <     * making decisions about creating and suspending spare
221 <     * threads. Updated only by CAS.  Note: CASes in
222 <     * updateRunningCount and preJoin assume that running active count
223 <     * is in low word, so need to be modified if this changes.
557 >     * The uncaught exception handler used when any worker abruptly
558 >     * terminates.
559       */
560 <    private volatile int workerCounts;
560 >    private final Thread.UncaughtExceptionHandler ueh;
561 >
562 >    /**
563 >     * Pool number, just for assigning useful names to worker threads
564 >     */
565 >    private final int poolNumber;
566 >
567  
568 <    private static int totalCountOf(int s)           { return s >>> 16;  }
569 <    private static int runningCountOf(int s)         { return s & shortMask; }
229 <    private static int workerCountsFor(int t, int r) { return (t << 16) + r; }
568 >    // Utilities for CASing fields. Note that several of these
569 >    // are manually inlined by callers
570  
571      /**
572 <     * Adds delta (which may be negative) to running count.  This must
233 <     * be called before (with negative arg) and after (with positive)
234 <     * any managed synchronization (i.e., mainly, joins).
235 <     * @param delta the number to add
572 >     * Increments running count part of workerCounts
573       */
574 <    final void updateRunningCount(int delta) {
575 <        int s;
576 <        do;while (!casWorkerCounts(s = workerCounts, s + delta));
574 >    final void incrementRunningCount() {
575 >        int c;
576 >        do {} while (!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
577 >                                               c = workerCounts,
578 >                                               c + ONE_RUNNING));
579      }
580  
581      /**
582 <     * Adds delta (which may be negative) to both total and running
244 <     * count.  This must be called upon creation and termination of
245 <     * worker threads.
246 <     * @param delta the number to add
582 >     * Tries to decrement running count unless already zero
583       */
584 <    private void updateWorkerCount(int delta) {
585 <        int d = delta + (delta << 16); // add to both lo and hi parts
586 <        int s;
587 <        do;while (!casWorkerCounts(s = workerCounts, s + d));
584 >    final boolean tryDecrementRunningCount() {
585 >        int wc = workerCounts;
586 >        if ((wc & RUNNING_COUNT_MASK) == 0)
587 >            return false;
588 >        return UNSAFE.compareAndSwapInt(this, workerCountsOffset,
589 >                                        wc, wc - ONE_RUNNING);
590      }
591  
592      /**
593 <     * Lifecycle control. High word contains runState, low word
594 <     * contains the number of workers that are (probably) executing
595 <     * tasks. This value is atomically incremented before a worker
596 <     * gets a task to run, and decremented when worker has no tasks
597 <     * and cannot find any. These two fields are bundled together to
260 <     * support correct termination triggering.  Note: activeCount
261 <     * CAS'es cheat by assuming active count is in low word, so need
262 <     * to be modified if this changes
593 >     * Forces decrement of encoded workerCounts, awaiting nonzero if
594 >     * (rarely) necessary when other count updates lag.
595 >     *
596 >     * @param dr -- either zero or ONE_RUNNING
597 >     * @param dt == either zero or ONE_TOTAL
598       */
599 <    private volatile int runControl;
600 <
601 <    // RunState values. Order among values matters
602 <    private static final int RUNNING     = 0;
603 <    private static final int SHUTDOWN    = 1;
604 <    private static final int TERMINATING = 2;
605 <    private static final int TERMINATED  = 3;
599 >    private void decrementWorkerCounts(int dr, int dt) {
600 >        for (;;) {
601 >            int wc = workerCounts;
602 >            if (wc == 0 && (runState & TERMINATED) != 0)
603 >                return; // lagging termination on a backout
604 >            if ((wc & RUNNING_COUNT_MASK)  - dr < 0 ||
605 >                (wc >>> TOTAL_COUNT_SHIFT) - dt < 0)
606 >                Thread.yield();
607 >            if (UNSAFE.compareAndSwapInt(this, workerCountsOffset,
608 >                                         wc, wc - (dr + dt)))
609 >                return;
610 >        }
611 >    }
612  
613 <    private static int runStateOf(int c)             { return c >>> 16; }
614 <    private static int activeCountOf(int c)          { return c & shortMask; }
615 <    private static int runControlFor(int r, int a)   { return (r << 16) + a; }
613 >    /**
614 >     * Increments event count
615 >     */
616 >    private void advanceEventCount() {
617 >        int c;
618 >        do {} while(!UNSAFE.compareAndSwapInt(this, eventCountOffset,
619 >                                              c = eventCount, c+1));
620 >    }
621  
622      /**
623 <     * Try incrementing active count; fail on contention. Called by
624 <     * workers before/during executing tasks.
623 >     * Tries incrementing active count; fails on contention.
624 >     * Called by workers before executing tasks.
625 >     *
626       * @return true on success
627       */
628      final boolean tryIncrementActiveCount() {
629 <        int c = runControl;
630 <        return casRunControl(c, c+1);
629 >        int c;
630 >        return UNSAFE.compareAndSwapInt(this, runStateOffset,
631 >                                        c = runState, c + ONE_ACTIVE);
632      }
633  
634      /**
635       * Tries decrementing active count; fails on contention.
636 <     * Possibly triggers termination on success.
289 <     * Called by workers when they can't find tasks.
290 <     * @return true on success
636 >     * Called when workers cannot find tasks to run.
637       */
638      final boolean tryDecrementActiveCount() {
639 <        int c = runControl;
640 <        int nextc = c - 1;
641 <        if (!casRunControl(c, nextc))
296 <            return false;
297 <        if (canTerminateOnShutdown(nextc))
298 <            terminateOnShutdown();
299 <        return true;
639 >        int c;
640 >        return UNSAFE.compareAndSwapInt(this, runStateOffset,
641 >                                        c = runState, c - ONE_ACTIVE);
642      }
643  
644      /**
645 <     * Returns true if argument represents zero active count and
646 <     * nonzero runstate, which is the triggering condition for
305 <     * terminating on shutdown.
645 >     * Advances to at least the given level. Returns true if not
646 >     * already in at least the given level.
647       */
648 <    private static boolean canTerminateOnShutdown(int c) {
649 <        return ((c & -c) >>> 16) != 0; // i.e. least bit is nonzero runState bit
648 >    private boolean advanceRunLevel(int level) {
649 >        for (;;) {
650 >            int s = runState;
651 >            if ((s & level) != 0)
652 >                return false;
653 >            if (UNSAFE.compareAndSwapInt(this, runStateOffset, s, s | level))
654 >                return true;
655 >        }
656      }
657  
658 +    // workers array maintenance
659 +
660      /**
661 <     * Transition run state to at least the given state. Return true
313 <     * if not already at least given state.
661 >     * Records and returns a workers array index for new worker.
662       */
663 <    private boolean transitionRunStateTo(int state) {
664 <        for (;;) {
665 <            int c = runControl;
666 <            if (runStateOf(c) >= state)
667 <                return false;
668 <            if (casRunControl(c, runControlFor(state, activeCountOf(c))))
669 <                return true;
663 >    private int recordWorker(ForkJoinWorkerThread w) {
664 >        // Try using slot totalCount-1. If not available, scan and/or resize
665 >        int k = (workerCounts >>> TOTAL_COUNT_SHIFT) - 1;
666 >        final ReentrantLock lock = this.workerLock;
667 >        lock.lock();
668 >        try {
669 >            ForkJoinWorkerThread[] ws = workers;
670 >            int n = ws.length;
671 >            if (k < 0 || k >= n || ws[k] != null) {
672 >                for (k = 0; k < n && ws[k] != null; ++k)
673 >                    ;
674 >                if (k == n)
675 >                    ws = Arrays.copyOf(ws, n << 1);
676 >            }
677 >            ws[k] = w;
678 >            workers = ws; // volatile array write ensures slot visibility
679 >        } finally {
680 >            lock.unlock();
681          }
682 +        return k;
683      }
684  
685      /**
686 <     * Controls whether to add spares to maintain parallelism
686 >     * Nulls out record of worker in workers array
687       */
688 <    private volatile boolean maintainsParallelism;
688 >    private void forgetWorker(ForkJoinWorkerThread w) {
689 >        int idx = w.poolIndex;
690 >        // Locking helps method recordWorker avoid unecessary expansion
691 >        final ReentrantLock lock = this.workerLock;
692 >        lock.lock();
693 >        try {
694 >            ForkJoinWorkerThread[] ws = workers;
695 >            if (idx >= 0 && idx < ws.length && ws[idx] == w) // verify
696 >                ws[idx] = null;
697 >        } finally {
698 >            lock.unlock();
699 >        }
700 >    }
701  
702 <    // Constructors
702 >    // adding and removing workers
703  
704      /**
705 <     * Creates a ForkJoinPool with a pool size equal to the number of
706 <     * processors available on the system and using the default
707 <     * ForkJoinWorkerThreadFactory,
336 <     * @throws SecurityException if a security manager exists and
337 <     *         the caller is not permitted to modify threads
338 <     *         because it does not hold {@link
339 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
705 >     * Tries to create and add new worker. Assumes that worker counts
706 >     * are already updated to accommodate the worker, so adjusts on
707 >     * failure.
708       */
709 <    public ForkJoinPool() {
710 <        this(Runtime.getRuntime().availableProcessors(),
711 <             defaultForkJoinWorkerThreadFactory);
709 >    private void addWorker() {
710 >        ForkJoinWorkerThread w = null;
711 >        try {
712 >            w = factory.newThread(this);
713 >        } finally { // Adjust on either null or exceptional factory return
714 >            if (w == null) {
715 >                decrementWorkerCounts(ONE_RUNNING, ONE_TOTAL);
716 >                tryTerminate(false); // in case of failure during shutdown
717 >            }
718 >        }
719 >        if (w != null)
720 >            w.start(recordWorker(w), ueh);
721      }
722  
723      /**
724 <     * Creates a ForkJoinPool with the indicated parallelism level
725 <     * threads, and using the default ForkJoinWorkerThreadFactory,
726 <     * @param parallelism the number of worker threads
727 <     * @throws IllegalArgumentException if parallelism less than or
728 <     * equal to zero
352 <     * @throws SecurityException if a security manager exists and
353 <     *         the caller is not permitted to modify threads
354 <     *         because it does not hold {@link
355 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
724 >     * Final callback from terminating worker.  Removes record of
725 >     * worker from array, and adjusts counts. If pool is shutting
726 >     * down, tries to complete terminatation.
727 >     *
728 >     * @param w the worker
729       */
730 <    public ForkJoinPool(int parallelism) {
731 <        this(parallelism, defaultForkJoinWorkerThreadFactory);
730 >    final void workerTerminated(ForkJoinWorkerThread w) {
731 >        forgetWorker(w);
732 >        decrementWorkerCounts(w.isTrimmed()? 0 : ONE_RUNNING, ONE_TOTAL);
733 >        while (w.stealCount != 0) // collect final count
734 >            tryAccumulateStealCount(w);
735 >        tryTerminate(false);
736      }
737  
738 +    // Waiting for and signalling events
739 +
740      /**
741 <     * Creates a ForkJoinPool with parallelism equal to the number of
742 <     * processors available on the system and using the given
743 <     * ForkJoinWorkerThreadFactory,
744 <     * @param factory the factory for creating new threads
745 <     * @throws NullPointerException if factory is null
746 <     * @throws SecurityException if a security manager exists and
747 <     *         the caller is not permitted to modify threads
369 <     *         because it does not hold {@link
370 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
741 >     * Releases workers blocked on a count not equal to current count.
742 >     * Normally called after precheck that eventWaiters isn't zero to
743 >     * avoid wasted array checks.
744 >     *
745 >     * @param signalling true if caller is a signalling worker so can
746 >     * exit upon (conservatively) detected contention by other threads
747 >     * who will continue to release
748       */
749 <    public ForkJoinPool(ForkJoinWorkerThreadFactory factory) {
750 <        this(Runtime.getRuntime().availableProcessors(), factory);
749 >    private void releaseEventWaiters(boolean signalling) {
750 >        ForkJoinWorkerThread[] ws = workers;
751 >        int n = ws.length;
752 >        long h; // head of stack
753 >        ForkJoinWorkerThread w; int id, ec;
754 >        while ((id = ((int)((h = eventWaiters) & WAITER_ID_MASK)) - 1) >= 0 &&
755 >               (int)(h >>> EVENT_COUNT_SHIFT) != (ec = eventCount) &&
756 >               id < n && (w = ws[id]) != null) {
757 >            if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
758 >                                          h, h = w.nextWaiter))
759 >                LockSupport.unpark(w);
760 >            if (signalling && (eventCount != ec || eventWaiters != h))
761 >                break;
762 >        }
763      }
764  
765      /**
766 <     * Creates a ForkJoinPool with the given parallelism and factory.
767 <     *
379 <     * @param parallelism the targeted number of worker threads
380 <     * @param factory the factory for creating new threads
381 <     * @throws IllegalArgumentException if parallelism less than or
382 <     * equal to zero, or greater than implementation limit
383 <     * @throws NullPointerException if factory is null
384 <     * @throws SecurityException if a security manager exists and
385 <     *         the caller is not permitted to modify threads
386 <     *         because it does not hold {@link
387 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
766 >     * Tries to advance eventCount and releases waiters. Called only
767 >     * from workers.
768       */
769 <    public ForkJoinPool(int parallelism, ForkJoinWorkerThreadFactory factory) {
770 <        if (parallelism <= 0 || parallelism > MAX_THREADS)
771 <            throw new IllegalArgumentException();
772 <        if (factory == null)
773 <            throw new NullPointerException();
394 <        checkPermission();
395 <        this.factory = factory;
396 <        this.parallelism = parallelism;
397 <        this.maxPoolSize = MAX_THREADS;
398 <        this.maintainsParallelism = true;
399 <        this.poolNumber = poolNumberGenerator.incrementAndGet();
400 <        this.workerLock = new ReentrantLock();
401 <        this.termination = workerLock.newCondition();
402 <        this.stealCount = new AtomicLong();
403 <        this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
404 <        // worker array and workers are lazily constructed
769 >    final void signalWork() {
770 >        int c; // try to increment event count -- CAS failure OK
771 >        UNSAFE.compareAndSwapInt(this, eventCountOffset, c = eventCount, c+1);
772 >        if (eventWaiters != 0L)
773 >            releaseEventWaiters(true);
774      }
775  
776      /**
777 <     * Create new worker using factory.
778 <     * @param index the index to assign worker
779 <     * @return new worker, or null of factory failed
777 >     * Blocks worker until terminating or event count
778 >     * advances from last value held by worker
779 >     *
780 >     * @param w the calling worker thread
781       */
782 <    private ForkJoinWorkerThread createWorker(int index) {
783 <        Thread.UncaughtExceptionHandler h = ueh;
784 <        ForkJoinWorkerThread w = factory.newThread(this);
785 <        if (w != null) {
786 <            w.poolIndex = index;
787 <            w.setDaemon(true);
788 <            w.setAsyncMode(locallyFifo);
789 <            w.setName("ForkJoinPool-" + poolNumber + "-worker-" + index);
790 <            if (h != null)
791 <                w.setUncaughtExceptionHandler(h);
782 >    private void eventSync(ForkJoinWorkerThread w) {
783 >        int wec = w.lastEventCount;
784 >        long nh = (((long)wec) << EVENT_COUNT_SHIFT) | ((long)(w.poolIndex+1));
785 >        long h;
786 >        while ((runState < SHUTDOWN || !tryTerminate(false)) &&
787 >               ((h = eventWaiters) == 0L ||
788 >                (int)(h >>> EVENT_COUNT_SHIFT) == wec) &&
789 >               eventCount == wec) {
790 >            if (UNSAFE.compareAndSwapLong(this, eventWaitersOffset,
791 >                                          w.nextWaiter = h, nh)) {
792 >                while (runState < TERMINATING && eventCount == wec) {
793 >                    if (!tryAccumulateStealCount(w))  // transfer while idle
794 >                        continue;
795 >                    Thread.interrupted();             // clear/ignore interrupt
796 >                    if (eventCount != wec)
797 >                        break;
798 >                    LockSupport.park(w);
799 >                }
800 >                break;
801 >            }
802          }
803 <        return w;
803 >        w.lastEventCount = eventCount;
804      }
805  
806 +    // Maintaining spares
807 +
808      /**
809 <     * Returns a good size for worker array given pool size.
428 <     * Currently requires size to be a power of two.
809 >     * Pushes worker onto the spare stack
810       */
811 <    private static int arraySizeFor(int ps) {
812 <        return ps <= 1? 1 : (1 << (32 - Integer.numberOfLeadingZeros(ps-1)));
811 >    final void pushSpare(ForkJoinWorkerThread w) {
812 >        int ns = (++w.spareCount << SPARE_COUNT_SHIFT) | (w.poolIndex+1);
813 >        do {} while (!UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
814 >                                               w.nextSpare = spareWaiters,ns));
815      }
816  
817      /**
818 <     * Creates or resizes array if necessary to hold newLength.
819 <     * Call only under exclusion.
437 <     *
438 <     * @return the array
818 >     * Tries (once) to resume a spare if running count is less than
819 >     * target parallelism. Fails on contention or stale workers.
820       */
821 <    private ForkJoinWorkerThread[] ensureWorkerArrayCapacity(int newLength) {
822 <        ForkJoinWorkerThread[] ws = workers;
823 <        if (ws == null)
824 <            return workers = new ForkJoinWorkerThread[arraySizeFor(newLength)];
825 <        else if (newLength > ws.length)
826 <            return workers = Arrays.copyOf(ws, arraySizeFor(newLength));
827 <        else
828 <            return ws;
821 >    private void tryResumeSpare() {
822 >        int sw, id;
823 >        ForkJoinWorkerThread w;
824 >        ForkJoinWorkerThread[] ws;
825 >        if ((id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 &&
826 >            id < (ws = workers).length && (w = ws[id]) != null &&
827 >            (workerCounts & RUNNING_COUNT_MASK) < parallelism &&
828 >            eventWaiters == 0L &&
829 >            spareWaiters == sw &&
830 >            UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
831 >                                     sw, w.nextSpare) &&
832 >            w.tryUnsuspend()) {
833 >            int c; // try increment; if contended, finish after unpark
834 >            boolean inc = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
835 >                                                   c = workerCounts,
836 >                                                   c + ONE_RUNNING);
837 >            LockSupport.unpark(w);
838 >            if (!inc) {
839 >                do {} while(!UNSAFE.compareAndSwapInt(this, workerCountsOffset,
840 >                                                      c = workerCounts,
841 >                                                      c + ONE_RUNNING));
842 >            }
843 >        }
844      }
845  
846      /**
847 <     * Try to shrink workers into smaller array after one or more terminate
847 >     * Callback from oldest spare occasionally waking up.  Tries
848 >     * (once) to shutdown a spare if more than 25% spare overage, or
849 >     * if UNUSED_SPARE_TRIM_RATE_NANOS have elapsed and there are at
850 >     * least #parallelism running threads. Note that we don't need CAS
851 >     * or locks here because the method is called only from the oldest
852 >     * suspended spare occasionally waking (and even misfires are OK).
853 >     *
854 >     * @param now the wake up nanoTime of caller
855       */
856 <    private void tryShrinkWorkerArray() {
857 <        ForkJoinWorkerThread[] ws = workers;
858 <        if (ws != null) {
859 <            int len = ws.length;
860 <            int last = len - 1;
861 <            while (last >= 0 && ws[last] == null)
862 <                --last;
863 <            int newLength = arraySizeFor(last+1);
864 <            if (newLength < len)
865 <                workers = Arrays.copyOf(ws, newLength);
856 >    final void tryTrimSpare(long now) {
857 >        long lastTrim = trimTime;
858 >        trimTime = now;
859 >        helpMaintainParallelism(); // first, help wake up any needed spares
860 >        int sw, id;
861 >        ForkJoinWorkerThread w;
862 >        ForkJoinWorkerThread[] ws;
863 >        int pc = parallelism;
864 >        int wc = workerCounts;
865 >        if ((wc & RUNNING_COUNT_MASK) >= pc &&
866 >            (((wc >>> TOTAL_COUNT_SHIFT) - pc) > (pc >>> 2) + 1 ||// approx 25%
867 >             now - lastTrim >= UNUSED_SPARE_TRIM_RATE_NANOS) &&
868 >            (id = ((sw = spareWaiters) & SPARE_ID_MASK) - 1) >= 0 &&
869 >            id < (ws = workers).length && (w = ws[id]) != null &&
870 >            UNSAFE.compareAndSwapInt(this, spareWaitersOffset,
871 >                                     sw, w.nextSpare))
872 >            w.shutdown(false);
873 >    }
874 >
875 >    /**
876 >     * Does at most one of:
877 >     *
878 >     * 1. Help wake up existing workers waiting for work via
879 >     *    releaseEventWaiters. (If any exist, then it probably doesn't
880 >     *    matter right now if under target parallelism level.)
881 >     *
882 >     * 2. If below parallelism level and a spare exists, try (once)
883 >     *    to resume it via tryResumeSpare.
884 >     *
885 >     * 3. If neither of the above, tries (once) to add a new
886 >     *    worker if either there are not enough total, or if all
887 >     *    existing workers are busy, there are either no running
888 >     *    workers or the deficit is at least twice the surplus.
889 >     */
890 >    private void helpMaintainParallelism() {
891 >        // uglified to work better when not compiled
892 >        int pc, wc, rc, tc, rs; long h;
893 >        if ((h = eventWaiters) != 0L) {
894 >            if ((int)(h >>> EVENT_COUNT_SHIFT) != eventCount)
895 >                releaseEventWaiters(false); // avoid useless call
896 >        }
897 >        else if ((pc = parallelism) >
898 >                 (rc = ((wc = workerCounts) & RUNNING_COUNT_MASK))) {
899 >            if (spareWaiters != 0)
900 >                tryResumeSpare();
901 >            else if ((rs = runState) < TERMINATING &&
902 >                     ((tc = wc >>> TOTAL_COUNT_SHIFT) < pc ||
903 >                      (tc == (rs & ACTIVE_COUNT_MASK) && // all busy
904 >                       (rc == 0 ||                       // must add
905 >                        rc < pc - ((tc - pc) << 1)) &&   // within slack
906 >                       tc < MAX_WORKERS && runState == rs)) && // recheck busy
907 >                     workerCounts == wc &&
908 >                     UNSAFE.compareAndSwapInt(this, workerCountsOffset, wc,
909 >                                              wc + (ONE_RUNNING|ONE_TOTAL)))
910 >                addWorker();
911 >        }
912 >    }
913 >
914 >    /**
915 >     * Callback from workers invoked upon each top-level action (i.e.,
916 >     * stealing a task or taking a submission and running
917 >     * it). Performs one or more of the following:
918 >     *
919 >     * 1. If the worker cannot find work (misses > 0), updates its
920 >     *    active status to inactive and updates activeCount unless
921 >     *    this is the first miss and there is contention, in which
922 >     *    case it may try again (either in this or a subsequent
923 >     *    call).
924 >     *
925 >     * 2. If there are at least 2 misses, awaits the next task event
926 >     *    via eventSync
927 >     *
928 >     * 3. If there are too many running threads, suspends this worker
929 >     *    (first forcing inactivation if necessary).  If it is not
930 >     *    needed, it may be killed while suspended via
931 >     *    tryTrimSpare. Otherwise, upon resume it rechecks to make
932 >     *    sure that it is still needed.
933 >     *
934 >     * 4. Helps release and/or reactivate other workers via
935 >     *    helpMaintainParallelism
936 >     *
937 >     * @param w the worker
938 >     * @param misses the number of scans by caller failing to find work
939 >     * (saturating at 2 just to avoid wraparound)
940 >     */
941 >    final void preStep(ForkJoinWorkerThread w, int misses) {
942 >        boolean active = w.active;
943 >        int pc = parallelism;
944 >        for (;;) {
945 >            int wc = workerCounts;
946 >            int rc = wc & RUNNING_COUNT_MASK;
947 >            if (active && (misses > 0 || rc > pc)) {
948 >                int rs;                      // try inactivate
949 >                if (UNSAFE.compareAndSwapInt(this, runStateOffset,
950 >                                             rs = runState, rs - ONE_ACTIVE))
951 >                    active = w.active = false;
952 >                else if (misses > 1 || rc > pc ||
953 >                         (rs & ACTIVE_COUNT_MASK) >= pc)
954 >                    continue;                // force inactivate
955 >            }
956 >            if (misses > 1) {
957 >                misses = 0;                  // don't re-sync
958 >                eventSync(w);                // continue loop to recheck rc
959 >            }
960 >            else if (rc > pc) {
961 >                if (workerCounts == wc &&   // try to suspend as spare
962 >                    UNSAFE.compareAndSwapInt(this, workerCountsOffset,
963 >                                             wc, wc - ONE_RUNNING) &&
964 >                    !w.suspendAsSpare())    // false if killed
965 >                    break;
966 >            }
967 >            else {
968 >                if (rc < pc || eventWaiters != 0L)
969 >                    helpMaintainParallelism();
970 >                break;
971 >            }
972          }
973      }
974  
975      /**
976 <     * Initialize workers if necessary
977 <     */
978 <    final void ensureWorkerInitialization() {
979 <        ForkJoinWorkerThread[] ws = workers;
980 <        if (ws == null) {
981 <            final ReentrantLock lock = this.workerLock;
982 <            lock.lock();
983 <            try {
984 <                ws = workers;
985 <                if (ws == null) {
986 <                    int ps = parallelism;
987 <                    ws = ensureWorkerArrayCapacity(ps);
988 <                    for (int i = 0; i < ps; ++i) {
989 <                        ForkJoinWorkerThread w = createWorker(i);
990 <                        if (w != null) {
991 <                            ws[i] = w;
992 <                            w.start();
993 <                            updateWorkerCount(1);
994 <                        }
995 <                    }
996 <                }
997 <            } finally {
998 <                lock.unlock();
976 >     * Helps and/or blocks awaiting join of the given task.
977 >     * Alternates between helpJoinTask() and helpMaintainParallelism()
978 >     * as many times as there is a deficit in running count (or longer
979 >     * if running count would become zero), then blocks if task still
980 >     * not done.
981 >     *
982 >     * @param joinMe the task to join
983 >     */
984 >    final void awaitJoin(ForkJoinTask<?> joinMe, ForkJoinWorkerThread worker) {
985 >        int threshold = parallelism;         // descend blocking thresholds
986 >        while (joinMe.status >= 0) {
987 >            boolean block; int wc;
988 >            worker.helpJoinTask(joinMe);
989 >            if (joinMe.status < 0)
990 >                break;
991 >            if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) {
992 >                if (threshold > 0)
993 >                    --threshold;
994 >                else
995 >                    advanceEventCount(); // force release
996 >                block = false;
997 >            }
998 >            else
999 >                block = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1000 >                                                 wc, wc - ONE_RUNNING);
1001 >            helpMaintainParallelism();
1002 >            if (block) {
1003 >                int c;
1004 >                joinMe.internalAwaitDone();
1005 >                do {} while (!UNSAFE.compareAndSwapInt
1006 >                             (this, workerCountsOffset,
1007 >                              c = workerCounts, c + ONE_RUNNING));
1008 >                break;
1009              }
1010          }
1011      }
1012  
1013      /**
1014 <     * Worker creation and startup for threads added via setParallelism.
1014 >     * Same idea as awaitJoin, but no helping
1015       */
1016 <    private void createAndStartAddedWorkers() {
1017 <        resumeAllSpares();  // Allow spares to convert to nonspare
1018 <        int ps = parallelism;
1019 <        ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(ps);
1020 <        int len = ws.length;
1021 <        // Sweep through slots, to keep lowest indices most populated
1022 <        int k = 0;
1023 <        while (k < len) {
1024 <            if (ws[k] != null) {
1025 <                ++k;
1026 <                continue;
1016 >    final void awaitBlocker(ManagedBlocker blocker)
1017 >        throws InterruptedException {
1018 >        int threshold = parallelism;
1019 >        while (!blocker.isReleasable()) {
1020 >            boolean block; int wc;
1021 >            if (((wc = workerCounts) & RUNNING_COUNT_MASK) <= threshold) {
1022 >                if (threshold > 0)
1023 >                    --threshold;
1024 >                else
1025 >                    advanceEventCount();
1026 >                block = false;
1027              }
1028 <            int s = workerCounts;
1029 <            int tc = totalCountOf(s);
1030 <            int rc = runningCountOf(s);
1031 <            if (rc >= ps || tc >= ps)
1028 >            else
1029 >                block = UNSAFE.compareAndSwapInt(this, workerCountsOffset,
1030 >                                                 wc, wc - ONE_RUNNING);
1031 >            helpMaintainParallelism();
1032 >            if (block) {
1033 >                try {
1034 >                    do {} while (!blocker.isReleasable() && !blocker.block());
1035 >                } finally {
1036 >                    int c;
1037 >                    do {} while (!UNSAFE.compareAndSwapInt
1038 >                                 (this, workerCountsOffset,
1039 >                                  c = workerCounts, c + ONE_RUNNING));
1040 >                }
1041                  break;
1042 <            if (casWorkerCounts (s, workerCountsFor(tc+1, rc+1))) {
1043 <                ForkJoinWorkerThread w = createWorker(k);
1042 >            }
1043 >        }
1044 >    }
1045 >
1046 >    /**
1047 >     * Possibly initiates and/or completes termination.
1048 >     *
1049 >     * @param now if true, unconditionally terminate, else only
1050 >     * if shutdown and empty queue and no active workers
1051 >     * @return true if now terminating or terminated
1052 >     */
1053 >    private boolean tryTerminate(boolean now) {
1054 >        if (now)
1055 >            advanceRunLevel(SHUTDOWN); // ensure at least SHUTDOWN
1056 >        else if (runState < SHUTDOWN ||
1057 >                 !submissionQueue.isEmpty() ||
1058 >                 (runState & ACTIVE_COUNT_MASK) != 0)
1059 >            return false;
1060 >
1061 >        if (advanceRunLevel(TERMINATING))
1062 >            startTerminating();
1063 >
1064 >        // Finish now if all threads terminated; else in some subsequent call
1065 >        if ((workerCounts >>> TOTAL_COUNT_SHIFT) == 0) {
1066 >            advanceRunLevel(TERMINATED);
1067 >            termination.arrive();
1068 >        }
1069 >        return true;
1070 >    }
1071 >
1072 >    /**
1073 >     * Actions on transition to TERMINATING
1074 >     *
1075 >     * Runs up to four passes through workers: (0) shutting down each
1076 >     * quietly (without waking up if parked) to quickly spread
1077 >     * notifications without unnecessary bouncing around event queues
1078 >     * etc (1) wake up and help cancel tasks (2) interrupt (3) mop up
1079 >     * races with interrupted workers
1080 >     */
1081 >    private void startTerminating() {
1082 >        cancelSubmissions();
1083 >        for (int passes = 0; passes < 4 && workerCounts != 0; ++passes) {
1084 >            advanceEventCount();
1085 >            eventWaiters = 0L; // clobber lists
1086 >            spareWaiters = 0;
1087 >            ForkJoinWorkerThread[] ws = workers;
1088 >            int n = ws.length;
1089 >            for (int i = 0; i < n; ++i) {
1090 >                ForkJoinWorkerThread w = ws[i];
1091                  if (w != null) {
1092 <                    ws[k++] = w;
1093 <                    w.start();
1094 <                }
1095 <                else {
1096 <                    updateWorkerCount(-1); // back out on failed creation
1097 <                    break;
1092 >                    w.shutdown(true);
1093 >                    if (passes > 0 && !w.isTerminated()) {
1094 >                        w.cancelTasks();
1095 >                        LockSupport.unpark(w);
1096 >                        if (passes > 1) {
1097 >                            try {
1098 >                                w.interrupt();
1099 >                            } catch (SecurityException ignore) {
1100 >                            }
1101 >                        }
1102 >                    }
1103                  }
1104              }
1105          }
1106      }
1107  
1108 +    /**
1109 +     * Clear out and cancel submissions, ignoring exceptions
1110 +     */
1111 +    private void cancelSubmissions() {
1112 +        ForkJoinTask<?> task;
1113 +        while ((task = submissionQueue.poll()) != null) {
1114 +            try {
1115 +                task.cancel(false);
1116 +            } catch (Throwable ignore) {
1117 +            }
1118 +        }
1119 +    }
1120 +
1121 +    // misc support for ForkJoinWorkerThread
1122 +
1123 +    /**
1124 +     * Returns pool number
1125 +     */
1126 +    final int getPoolNumber() {
1127 +        return poolNumber;
1128 +    }
1129 +
1130 +    /**
1131 +     * Tries to accumulates steal count from a worker, clearing
1132 +     * the worker's value.
1133 +     *
1134 +     * @return true if worker steal count now zero
1135 +     */
1136 +    final boolean tryAccumulateStealCount(ForkJoinWorkerThread w) {
1137 +        int sc = w.stealCount;
1138 +        long c = stealCount;
1139 +        // CAS even if zero, for fence effects
1140 +        if (UNSAFE.compareAndSwapLong(this, stealCountOffset, c, c + sc)) {
1141 +            if (sc != 0)
1142 +                w.stealCount = 0;
1143 +            return true;
1144 +        }
1145 +        return sc == 0;
1146 +    }
1147 +
1148 +    /**
1149 +     * Returns the approximate (non-atomic) number of idle threads per
1150 +     * active thread.
1151 +     */
1152 +    final int idlePerActive() {
1153 +        int pc = parallelism; // use parallelism, not rc
1154 +        int ac = runState;    // no mask -- artifically boosts during shutdown
1155 +        // Use exact results for small values, saturate past 4
1156 +        return pc <= ac? 0 : pc >>> 1 <= ac? 1 : pc >>> 2 <= ac? 3 : pc >>> 3;
1157 +    }
1158 +
1159 +    // Public and protected methods
1160 +
1161 +    // Constructors
1162 +
1163 +    /**
1164 +     * Creates a {@code ForkJoinPool} with parallelism equal to {@link
1165 +     * java.lang.Runtime#availableProcessors}, using the {@linkplain
1166 +     * #defaultForkJoinWorkerThreadFactory default thread factory},
1167 +     * no UncaughtExceptionHandler, and non-async LIFO processing mode.
1168 +     *
1169 +     * @throws SecurityException if a security manager exists and
1170 +     *         the caller is not permitted to modify threads
1171 +     *         because it does not hold {@link
1172 +     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1173 +     */
1174 +    public ForkJoinPool() {
1175 +        this(Runtime.getRuntime().availableProcessors(),
1176 +             defaultForkJoinWorkerThreadFactory, null, false);
1177 +    }
1178 +
1179 +    /**
1180 +     * Creates a {@code ForkJoinPool} with the indicated parallelism
1181 +     * level, the {@linkplain
1182 +     * #defaultForkJoinWorkerThreadFactory default thread factory},
1183 +     * no UncaughtExceptionHandler, and non-async LIFO processing mode.
1184 +     *
1185 +     * @param parallelism the parallelism level
1186 +     * @throws IllegalArgumentException if parallelism less than or
1187 +     *         equal to zero, or greater than implementation limit
1188 +     * @throws SecurityException if a security manager exists and
1189 +     *         the caller is not permitted to modify threads
1190 +     *         because it does not hold {@link
1191 +     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1192 +     */
1193 +    public ForkJoinPool(int parallelism) {
1194 +        this(parallelism, defaultForkJoinWorkerThreadFactory, null, false);
1195 +    }
1196 +
1197 +    /**
1198 +     * Creates a {@code ForkJoinPool} with the given parameters.
1199 +     *
1200 +     * @param parallelism the parallelism level. For default value,
1201 +     * use {@link java.lang.Runtime#availableProcessors}.
1202 +     * @param factory the factory for creating new threads. For default value,
1203 +     * use {@link #defaultForkJoinWorkerThreadFactory}.
1204 +     * @param handler the handler for internal worker threads that
1205 +     * terminate due to unrecoverable errors encountered while executing
1206 +     * tasks. For default value, use <code>null</code>.
1207 +     * @param asyncMode if true,
1208 +     * establishes local first-in-first-out scheduling mode for forked
1209 +     * tasks that are never joined. This mode may be more appropriate
1210 +     * than default locally stack-based mode in applications in which
1211 +     * worker threads only process event-style asynchronous tasks.
1212 +     * For default value, use <code>false</code>.
1213 +     * @throws IllegalArgumentException if parallelism less than or
1214 +     *         equal to zero, or greater than implementation limit
1215 +     * @throws NullPointerException if the factory is null
1216 +     * @throws SecurityException if a security manager exists and
1217 +     *         the caller is not permitted to modify threads
1218 +     *         because it does not hold {@link
1219 +     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1220 +     */
1221 +    public ForkJoinPool(int parallelism,
1222 +                        ForkJoinWorkerThreadFactory factory,
1223 +                        Thread.UncaughtExceptionHandler handler,
1224 +                        boolean asyncMode) {
1225 +        checkPermission();
1226 +        if (factory == null)
1227 +            throw new NullPointerException();
1228 +        if (parallelism <= 0 || parallelism > MAX_WORKERS)
1229 +            throw new IllegalArgumentException();
1230 +        this.parallelism = parallelism;
1231 +        this.factory = factory;
1232 +        this.ueh = handler;
1233 +        this.locallyFifo = asyncMode;
1234 +        int arraySize = initialArraySizeFor(parallelism);
1235 +        this.workers = new ForkJoinWorkerThread[arraySize];
1236 +        this.submissionQueue = new LinkedTransferQueue<ForkJoinTask<?>>();
1237 +        this.workerLock = new ReentrantLock();
1238 +        this.termination = new Phaser(1);
1239 +        this.poolNumber = poolNumberGenerator.incrementAndGet();
1240 +        this.trimTime = System.nanoTime();
1241 +    }
1242 +
1243 +    /**
1244 +     * Returns initial power of two size for workers array.
1245 +     * @param pc the initial parallelism level
1246 +     */
1247 +    private static int initialArraySizeFor(int pc) {
1248 +        // See Hackers Delight, sec 3.2. We know MAX_WORKERS < (1 >>> 16)
1249 +        int size = pc < MAX_WORKERS ? pc + 1 : MAX_WORKERS;
1250 +        size |= size >>> 1;
1251 +        size |= size >>> 2;
1252 +        size |= size >>> 4;
1253 +        size |= size >>> 8;
1254 +        return size + 1;
1255 +    }
1256 +
1257      // Execution methods
1258  
1259      /**
1260       * Common code for execute, invoke and submit
1261       */
1262      private <T> void doSubmit(ForkJoinTask<T> task) {
1263 <        if (isShutdown())
1263 >        if (task == null)
1264 >            throw new NullPointerException();
1265 >        if (runState >= SHUTDOWN)
1266              throw new RejectedExecutionException();
536        if (workers == null)
537            ensureWorkerInitialization();
1267          submissionQueue.offer(task);
1268 <        signalIdleWorkers();
1268 >        advanceEventCount();
1269 >        helpMaintainParallelism();         // start or wake up workers
1270      }
1271  
1272      /**
1273 <     * Performs the given task; returning its result upon completion
1273 >     * Performs the given task, returning its result upon completion.
1274 >     *
1275       * @param task the task
1276       * @return the task's result
1277 <     * @throws NullPointerException if task is null
1278 <     * @throws RejectedExecutionException if pool is shut down
1277 >     * @throws NullPointerException if the task is null
1278 >     * @throws RejectedExecutionException if the task cannot be
1279 >     *         scheduled for execution
1280       */
1281      public <T> T invoke(ForkJoinTask<T> task) {
1282          doSubmit(task);
# Line 553 | Line 1285 | public class ForkJoinPool extends Abstra
1285  
1286      /**
1287       * Arranges for (asynchronous) execution of the given task.
1288 +     * If the caller is already engaged in a fork/join computation in
1289 +     * the current pool, this method is equivalent in effect to
1290 +     * {@link ForkJoinTask#fork}.
1291 +     *
1292       * @param task the task
1293 <     * @throws NullPointerException if task is null
1294 <     * @throws RejectedExecutionException if pool is shut down
1293 >     * @throws NullPointerException if the task is null
1294 >     * @throws RejectedExecutionException if the task cannot be
1295 >     *         scheduled for execution
1296       */
1297 <    public <T> void execute(ForkJoinTask<T> task) {
1297 >    public void execute(ForkJoinTask<?> task) {
1298          doSubmit(task);
1299      }
1300  
1301      // AbstractExecutorService methods
1302  
1303 +    /**
1304 +     * @throws NullPointerException if the task is null
1305 +     * @throws RejectedExecutionException if the task cannot be
1306 +     *         scheduled for execution
1307 +     */
1308      public void execute(Runnable task) {
1309 <        doSubmit(new AdaptedRunnable<Void>(task, null));
1309 >        ForkJoinTask<?> job;
1310 >        if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1311 >            job = (ForkJoinTask<?>) task;
1312 >        else
1313 >            job = ForkJoinTask.adapt(task, null);
1314 >        doSubmit(job);
1315      }
1316  
1317 +    /**
1318 +     * Submits a ForkJoinTask for execution.
1319 +     * If the caller is already engaged in a fork/join computation in
1320 +     * the current pool, this method is equivalent in effect to
1321 +     * {@link ForkJoinTask#fork}.
1322 +     *
1323 +     * @param task the task to submit
1324 +     * @return the task
1325 +     * @throws NullPointerException if the task is null
1326 +     * @throws RejectedExecutionException if the task cannot be
1327 +     *         scheduled for execution
1328 +     */
1329 +    public <T> ForkJoinTask<T> submit(ForkJoinTask<T> task) {
1330 +        doSubmit(task);
1331 +        return task;
1332 +    }
1333 +
1334 +    /**
1335 +     * @throws NullPointerException if the task is null
1336 +     * @throws RejectedExecutionException if the task cannot be
1337 +     *         scheduled for execution
1338 +     */
1339      public <T> ForkJoinTask<T> submit(Callable<T> task) {
1340 <        ForkJoinTask<T> job = new AdaptedCallable<T>(task);
1340 >        ForkJoinTask<T> job = ForkJoinTask.adapt(task);
1341          doSubmit(job);
1342          return job;
1343      }
1344  
1345 +    /**
1346 +     * @throws NullPointerException if the task is null
1347 +     * @throws RejectedExecutionException if the task cannot be
1348 +     *         scheduled for execution
1349 +     */
1350      public <T> ForkJoinTask<T> submit(Runnable task, T result) {
1351 <        ForkJoinTask<T> job = new AdaptedRunnable<T>(task, result);
1351 >        ForkJoinTask<T> job = ForkJoinTask.adapt(task, result);
1352          doSubmit(job);
1353          return job;
1354      }
1355  
1356 +    /**
1357 +     * @throws NullPointerException if the task is null
1358 +     * @throws RejectedExecutionException if the task cannot be
1359 +     *         scheduled for execution
1360 +     */
1361      public ForkJoinTask<?> submit(Runnable task) {
1362 <        ForkJoinTask<Void> job = new AdaptedRunnable<Void>(task, null);
1362 >        ForkJoinTask<?> job;
1363 >        if (task instanceof ForkJoinTask<?>) // avoid re-wrap
1364 >            job = (ForkJoinTask<?>) task;
1365 >        else
1366 >            job = ForkJoinTask.adapt(task, null);
1367          doSubmit(job);
1368          return job;
1369      }
1370  
1371      /**
1372 <     * Adaptor for Runnables. This implements RunnableFuture
1373 <     * to be compliant with AbstractExecutorService constraints
1372 >     * @throws NullPointerException       {@inheritDoc}
1373 >     * @throws RejectedExecutionException {@inheritDoc}
1374       */
592    static final class AdaptedRunnable<T> extends ForkJoinTask<T>
593        implements RunnableFuture<T> {
594        final Runnable runnable;
595        final T resultOnCompletion;
596        T result;
597        AdaptedRunnable(Runnable runnable, T result) {
598            if (runnable == null) throw new NullPointerException();
599            this.runnable = runnable;
600            this.resultOnCompletion = result;
601        }
602        public T getRawResult() { return result; }
603        public void setRawResult(T v) { result = v; }
604        public boolean exec() {
605            runnable.run();
606            result = resultOnCompletion;
607            return true;
608        }
609        public void run() { invoke(); }
610    }
611
612    /**
613     * Adaptor for Callables
614     */
615    static final class AdaptedCallable<T> extends ForkJoinTask<T>
616        implements RunnableFuture<T> {
617        final Callable<T> callable;
618        T result;
619        AdaptedCallable(Callable<T> callable) {
620            if (callable == null) throw new NullPointerException();
621            this.callable = callable;
622        }
623        public T getRawResult() { return result; }
624        public void setRawResult(T v) { result = v; }
625        public boolean exec() {
626            try {
627                result = callable.call();
628                return true;
629            } catch (Error err) {
630                throw err;
631            } catch (RuntimeException rex) {
632                throw rex;
633            } catch (Exception ex) {
634                throw new RuntimeException(ex);
635            }
636        }
637        public void run() { invoke(); }
638    }
639
1375      public <T> List<Future<T>> invokeAll(Collection<? extends Callable<T>> tasks) {
1376 <        ArrayList<ForkJoinTask<T>> ts =
1376 >        ArrayList<ForkJoinTask<T>> forkJoinTasks =
1377              new ArrayList<ForkJoinTask<T>>(tasks.size());
1378 <        for (Callable<T> c : tasks)
1379 <            ts.add(new AdaptedCallable<T>(c));
1380 <        invoke(new InvokeAll<T>(ts));
1381 <        return (List<Future<T>>)(List)ts;
1378 >        for (Callable<T> task : tasks)
1379 >            forkJoinTasks.add(ForkJoinTask.adapt(task));
1380 >        invoke(new InvokeAll<T>(forkJoinTasks));
1381 >
1382 >        @SuppressWarnings({"unchecked", "rawtypes"})
1383 >            List<Future<T>> futures = (List<Future<T>>) (List) forkJoinTasks;
1384 >        return futures;
1385      }
1386  
1387      static final class InvokeAll<T> extends RecursiveAction {
1388          final ArrayList<ForkJoinTask<T>> tasks;
1389          InvokeAll(ArrayList<ForkJoinTask<T>> tasks) { this.tasks = tasks; }
1390          public void compute() {
1391 <            try { invokeAll(tasks); } catch(Exception ignore) {}
1391 >            try { invokeAll(tasks); }
1392 >            catch (Exception ignore) {}
1393          }
1394 +        private static final long serialVersionUID = -7914297376763021607L;
1395      }
1396  
657    // Configuration and status settings and queries
658
1397      /**
1398 <     * Returns the factory used for constructing new workers
1398 >     * Returns the factory used for constructing new workers.
1399       *
1400       * @return the factory used for constructing new workers
1401       */
# Line 668 | Line 1406 | public class ForkJoinPool extends Abstra
1406      /**
1407       * Returns the handler for internal worker threads that terminate
1408       * due to unrecoverable errors encountered while executing tasks.
671     * @return the handler, or null if none
672     */
673    public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
674        Thread.UncaughtExceptionHandler h;
675        final ReentrantLock lock = this.workerLock;
676        lock.lock();
677        try {
678            h = ueh;
679        } finally {
680            lock.unlock();
681        }
682        return h;
683    }
684
685    /**
686     * Sets the handler for internal worker threads that terminate due
687     * to unrecoverable errors encountered while executing tasks.
688     * Unless set, the current default or ThreadGroup handler is used
689     * as handler.
1409       *
1410 <     * @param h the new handler
692 <     * @return the old handler, or null if none
693 <     * @throws SecurityException if a security manager exists and
694 <     *         the caller is not permitted to modify threads
695 <     *         because it does not hold {@link
696 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
697 <     */
698 <    public Thread.UncaughtExceptionHandler
699 <        setUncaughtExceptionHandler(Thread.UncaughtExceptionHandler h) {
700 <        checkPermission();
701 <        Thread.UncaughtExceptionHandler old = null;
702 <        final ReentrantLock lock = this.workerLock;
703 <        lock.lock();
704 <        try {
705 <            old = ueh;
706 <            ueh = h;
707 <            ForkJoinWorkerThread[] ws = workers;
708 <            if (ws != null) {
709 <                for (int i = 0; i < ws.length; ++i) {
710 <                    ForkJoinWorkerThread w = ws[i];
711 <                    if (w != null)
712 <                        w.setUncaughtExceptionHandler(h);
713 <                }
714 <            }
715 <        } finally {
716 <            lock.unlock();
717 <        }
718 <        return old;
719 <    }
720 <
721 <
722 <    /**
723 <     * Sets the target parallelism level of this pool.
724 <     * @param parallelism the target parallelism
725 <     * @throws IllegalArgumentException if parallelism less than or
726 <     * equal to zero or greater than maximum size bounds
727 <     * @throws SecurityException if a security manager exists and
728 <     *         the caller is not permitted to modify threads
729 <     *         because it does not hold {@link
730 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
1410 >     * @return the handler, or {@code null} if none
1411       */
1412 <    public void setParallelism(int parallelism) {
1413 <        checkPermission();
734 <        if (parallelism <= 0 || parallelism > maxPoolSize)
735 <            throw new IllegalArgumentException();
736 <        final ReentrantLock lock = this.workerLock;
737 <        lock.lock();
738 <        try {
739 <            if (!isTerminating()) {
740 <                int p = this.parallelism;
741 <                this.parallelism = parallelism;
742 <                if (parallelism > p)
743 <                    createAndStartAddedWorkers();
744 <                else
745 <                    trimSpares();
746 <            }
747 <        } finally {
748 <            lock.unlock();
749 <        }
750 <        signalIdleWorkers();
1412 >    public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
1413 >        return ueh;
1414      }
1415  
1416      /**
1417 <     * Returns the targeted number of worker threads in this pool.
1417 >     * Returns the targeted parallelism level of this pool.
1418       *
1419 <     * @return the targeted number of worker threads in this pool
1419 >     * @return the targeted parallelism level of this pool
1420       */
1421      public int getParallelism() {
1422          return parallelism;
# Line 762 | Line 1425 | public class ForkJoinPool extends Abstra
1425      /**
1426       * Returns the number of worker threads that have started but not
1427       * yet terminated.  This result returned by this method may differ
1428 <     * from {@code getParallelism} when threads are created to
1428 >     * from {@link #getParallelism} when threads are created to
1429       * maintain parallelism when others are cooperatively blocked.
1430       *
1431       * @return the number of worker threads
1432       */
1433      public int getPoolSize() {
1434 <        return totalCountOf(workerCounts);
772 <    }
773 <
774 <    /**
775 <     * Returns the maximum number of threads allowed to exist in the
776 <     * pool, even if there are insufficient unblocked running threads.
777 <     * @return the maximum
778 <     */
779 <    public int getMaximumPoolSize() {
780 <        return maxPoolSize;
781 <    }
782 <
783 <    /**
784 <     * Sets the maximum number of threads allowed to exist in the
785 <     * pool, even if there are insufficient unblocked running threads.
786 <     * Setting this value has no effect on current pool size. It
787 <     * controls construction of new threads.
788 <     * @throws IllegalArgumentException if negative or greater then
789 <     * internal implementation limit
790 <     */
791 <    public void setMaximumPoolSize(int newMax) {
792 <        if (newMax < 0 || newMax > MAX_THREADS)
793 <            throw new IllegalArgumentException();
794 <        maxPoolSize = newMax;
795 <    }
796 <
797 <
798 <    /**
799 <     * Returns true if this pool dynamically maintains its target
800 <     * parallelism level. If false, new threads are added only to
801 <     * avoid possible starvation.
802 <     * This setting is by default true;
803 <     * @return true if maintains parallelism
804 <     */
805 <    public boolean getMaintainsParallelism() {
806 <        return maintainsParallelism;
807 <    }
808 <
809 <    /**
810 <     * Sets whether this pool dynamically maintains its target
811 <     * parallelism level. If false, new threads are added only to
812 <     * avoid possible starvation.
813 <     * @param enable true to maintains parallelism
814 <     */
815 <    public void setMaintainsParallelism(boolean enable) {
816 <        maintainsParallelism = enable;
1434 >        return workerCounts >>> TOTAL_COUNT_SHIFT;
1435      }
1436  
1437      /**
1438 <     * Establishes local first-in-first-out scheduling mode for forked
821 <     * tasks that are never joined. This mode may be more appropriate
822 <     * than default locally stack-based mode in applications in which
823 <     * worker threads only process asynchronous tasks.  This method is
824 <     * designed to be invoked only when pool is quiescent, and
825 <     * typically only before any tasks are submitted. The effects of
826 <     * invocations at other times may be unpredictable.
827 <     *
828 <     * @param async if true, use locally FIFO scheduling
829 <     * @return the previous mode
830 <     */
831 <    public boolean setAsyncMode(boolean async) {
832 <        boolean oldMode = locallyFifo;
833 <        locallyFifo = async;
834 <        ForkJoinWorkerThread[] ws = workers;
835 <        if (ws != null) {
836 <            for (int i = 0; i < ws.length; ++i) {
837 <                ForkJoinWorkerThread t = ws[i];
838 <                if (t != null)
839 <                    t.setAsyncMode(async);
840 <            }
841 <        }
842 <        return oldMode;
843 <    }
844 <
845 <    /**
846 <     * Returns true if this pool uses local first-in-first-out
1438 >     * Returns {@code true} if this pool uses local first-in-first-out
1439       * scheduling mode for forked tasks that are never joined.
1440       *
1441 <     * @return true if this pool uses async mode
1441 >     * @return {@code true} if this pool uses async mode
1442       */
1443      public boolean getAsyncMode() {
1444          return locallyFifo;
# Line 855 | Line 1447 | public class ForkJoinPool extends Abstra
1447      /**
1448       * Returns an estimate of the number of worker threads that are
1449       * not blocked waiting to join tasks or for other managed
1450 <     * synchronization.
1450 >     * synchronization. This method may overestimate the
1451 >     * number of running threads.
1452       *
1453       * @return the number of worker threads
1454       */
1455      public int getRunningThreadCount() {
1456 <        return runningCountOf(workerCounts);
1456 >        return workerCounts & RUNNING_COUNT_MASK;
1457      }
1458  
1459      /**
1460       * Returns an estimate of the number of threads that are currently
1461       * stealing or executing tasks. This method may overestimate the
1462       * number of active threads.
1463 +     *
1464       * @return the number of active threads
1465       */
1466      public int getActiveThreadCount() {
1467 <        return activeCountOf(runControl);
874 <    }
875 <
876 <    /**
877 <     * Returns an estimate of the number of threads that are currently
878 <     * idle waiting for tasks. This method may underestimate the
879 <     * number of idle threads.
880 <     * @return the number of idle threads
881 <     */
882 <    final int getIdleThreadCount() {
883 <        int c = runningCountOf(workerCounts) - activeCountOf(runControl);
884 <        return (c <= 0)? 0 : c;
1467 >        return runState & ACTIVE_COUNT_MASK;
1468      }
1469  
1470      /**
1471 <     * Returns true if all worker threads are currently idle. An idle
1472 <     * worker is one that cannot obtain a task to execute because none
1473 <     * are available to steal from other threads, and there are no
1474 <     * pending submissions to the pool. This method is conservative:
1475 <     * It might not return true immediately upon idleness of all
1476 <     * threads, but will eventually become true if threads remain
1477 <     * inactive.
1478 <     * @return true if all threads are currently idle
1471 >     * Returns {@code true} if all worker threads are currently idle.
1472 >     * An idle worker is one that cannot obtain a task to execute
1473 >     * because none are available to steal from other threads, and
1474 >     * there are no pending submissions to the pool. This method is
1475 >     * conservative; it might not return {@code true} immediately upon
1476 >     * idleness of all threads, but will eventually become true if
1477 >     * threads remain inactive.
1478 >     *
1479 >     * @return {@code true} if all threads are currently idle
1480       */
1481      public boolean isQuiescent() {
1482 <        return activeCountOf(runControl) == 0;
1482 >        return (runState & ACTIVE_COUNT_MASK) == 0;
1483      }
1484  
1485      /**
# Line 903 | Line 1487 | public class ForkJoinPool extends Abstra
1487       * one thread's work queue by another. The reported value
1488       * underestimates the actual total number of steals when the pool
1489       * is not quiescent. This value may be useful for monitoring and
1490 <     * tuning fork/join programs: In general, steal counts should be
1490 >     * tuning fork/join programs: in general, steal counts should be
1491       * high enough to keep threads busy, but low enough to avoid
1492       * overhead and contention across threads.
1493 +     *
1494       * @return the number of steals
1495       */
1496      public long getStealCount() {
1497 <        return stealCount.get();
913 <    }
914 <
915 <    /**
916 <     * Accumulate steal count from a worker. Call only
917 <     * when worker known to be idle.
918 <     */
919 <    private void updateStealCount(ForkJoinWorkerThread w) {
920 <        int sc = w.getAndClearStealCount();
921 <        if (sc != 0)
922 <            stealCount.addAndGet(sc);
1497 >        return stealCount;
1498      }
1499  
1500      /**
# Line 929 | Line 1504 | public class ForkJoinPool extends Abstra
1504       * an approximation, obtained by iterating across all threads in
1505       * the pool. This method may be useful for tuning task
1506       * granularities.
1507 +     *
1508       * @return the number of queued tasks
1509       */
1510      public long getQueuedTaskCount() {
1511          long count = 0;
1512          ForkJoinWorkerThread[] ws = workers;
1513 <        if (ws != null) {
1514 <            for (int i = 0; i < ws.length; ++i) {
1515 <                ForkJoinWorkerThread t = ws[i];
1516 <                if (t != null)
1517 <                    count += t.getQueueSize();
942 <            }
1513 >        int n = ws.length;
1514 >        for (int i = 0; i < n; ++i) {
1515 >            ForkJoinWorkerThread w = ws[i];
1516 >            if (w != null)
1517 >                count += w.getQueueSize();
1518          }
1519          return count;
1520      }
1521  
1522      /**
1523 <     * Returns an estimate of the number tasks submitted to this pool
1524 <     * that have not yet begun executing. This method takes time
1523 >     * Returns an estimate of the number of tasks submitted to this
1524 >     * pool that have not yet begun executing.  This method takes time
1525       * proportional to the number of submissions.
1526 +     *
1527       * @return the number of queued submissions
1528       */
1529      public int getQueuedSubmissionCount() {
# Line 955 | Line 1531 | public class ForkJoinPool extends Abstra
1531      }
1532  
1533      /**
1534 <     * Returns true if there are any tasks submitted to this pool
1535 <     * that have not yet begun executing.
1534 >     * Returns {@code true} if there are any tasks submitted to this
1535 >     * pool that have not yet begun executing.
1536 >     *
1537       * @return {@code true} if there are any queued submissions
1538       */
1539      public boolean hasQueuedSubmissions() {
# Line 967 | Line 1544 | public class ForkJoinPool extends Abstra
1544       * Removes and returns the next unexecuted submission if one is
1545       * available.  This method may be useful in extensions to this
1546       * class that re-assign work in systems with multiple pools.
1547 <     * @return the next submission, or null if none
1547 >     *
1548 >     * @return the next submission, or {@code null} if none
1549       */
1550      protected ForkJoinTask<?> pollSubmission() {
1551          return submissionQueue.poll();
# Line 977 | Line 1555 | public class ForkJoinPool extends Abstra
1555       * Removes all available unexecuted submitted and forked tasks
1556       * from scheduling queues and adds them to the given collection,
1557       * without altering their execution status. These may include
1558 <     * artificially generated or wrapped tasks. This method is designed
1559 <     * to be invoked only when the pool is known to be
1558 >     * artificially generated or wrapped tasks. This method is
1559 >     * designed to be invoked only when the pool is known to be
1560       * quiescent. Invocations at other times may not remove all
1561       * tasks. A failure encountered while attempting to add elements
1562       * to collection {@code c} may result in elements being in
# Line 986 | Line 1564 | public class ForkJoinPool extends Abstra
1564       * exception is thrown.  The behavior of this operation is
1565       * undefined if the specified collection is modified while the
1566       * operation is in progress.
1567 +     *
1568       * @param c the collection to transfer elements into
1569       * @return the number of elements transferred
1570       */
1571 <    protected int drainTasksTo(Collection<ForkJoinTask<?>> c) {
1572 <        int n = submissionQueue.drainTo(c);
1571 >    protected int drainTasksTo(Collection<? super ForkJoinTask<?>> c) {
1572 >        int count = submissionQueue.drainTo(c);
1573          ForkJoinWorkerThread[] ws = workers;
1574 <        if (ws != null) {
1575 <            for (int i = 0; i < ws.length; ++i) {
1576 <                ForkJoinWorkerThread w = ws[i];
1577 <                if (w != null)
1578 <                    n += w.drainTasksTo(c);
1000 <            }
1574 >        int n = ws.length;
1575 >        for (int i = 0; i < n; ++i) {
1576 >            ForkJoinWorkerThread w = ws[i];
1577 >            if (w != null)
1578 >                count += w.drainTasksTo(c);
1579          }
1580 <        return n;
1580 >        return count;
1581      }
1582  
1583      /**
# Line 1010 | Line 1588 | public class ForkJoinPool extends Abstra
1588       * @return a string identifying this pool, as well as its state
1589       */
1590      public String toString() {
1013        int ps = parallelism;
1014        int wc = workerCounts;
1015        int rc = runControl;
1591          long st = getStealCount();
1592          long qt = getQueuedTaskCount();
1593          long qs = getQueuedSubmissionCount();
1594 +        int wc = workerCounts;
1595 +        int tc = wc >>> TOTAL_COUNT_SHIFT;
1596 +        int rc = wc & RUNNING_COUNT_MASK;
1597 +        int pc = parallelism;
1598 +        int rs = runState;
1599 +        int ac = rs & ACTIVE_COUNT_MASK;
1600          return super.toString() +
1601 <            "[" + runStateToString(runStateOf(rc)) +
1602 <            ", parallelism = " + ps +
1603 <            ", size = " + totalCountOf(wc) +
1604 <            ", active = " + activeCountOf(rc) +
1605 <            ", running = " + runningCountOf(wc) +
1601 >            "[" + runLevelToString(rs) +
1602 >            ", parallelism = " + pc +
1603 >            ", size = " + tc +
1604 >            ", active = " + ac +
1605 >            ", running = " + rc +
1606              ", steals = " + st +
1607              ", tasks = " + qt +
1608              ", submissions = " + qs +
1609              "]";
1610      }
1611  
1612 <    private static String runStateToString(int rs) {
1613 <        switch(rs) {
1614 <        case RUNNING: return "Running";
1615 <        case SHUTDOWN: return "Shutting down";
1616 <        case TERMINATING: return "Terminating";
1036 <        case TERMINATED: return "Terminated";
1037 <        default: throw new Error("Unknown run state");
1038 <        }
1612 >    private static String runLevelToString(int s) {
1613 >        return ((s & TERMINATED) != 0 ? "Terminated" :
1614 >                ((s & TERMINATING) != 0 ? "Terminating" :
1615 >                 ((s & SHUTDOWN) != 0 ? "Shutting down" :
1616 >                  "Running")));
1617      }
1618  
1041    // lifecycle control
1042
1619      /**
1620       * Initiates an orderly shutdown in which previously submitted
1621       * tasks are executed, but no new tasks will be accepted.
1622       * Invocation has no additional effect if already shut down.
1623       * Tasks that are in the process of being submitted concurrently
1624       * during the course of this method may or may not be rejected.
1625 +     *
1626       * @throws SecurityException if a security manager exists and
1627       *         the caller is not permitted to modify threads
1628       *         because it does not hold {@link
1629 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
1629 >     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1630       */
1631      public void shutdown() {
1632          checkPermission();
1633 <        transitionRunStateTo(SHUTDOWN);
1634 <        if (canTerminateOnShutdown(runControl))
1058 <            terminateOnShutdown();
1633 >        advanceRunLevel(SHUTDOWN);
1634 >        tryTerminate(false);
1635      }
1636  
1637      /**
1638 <     * Attempts to stop all actively executing tasks, and cancels all
1639 <     * waiting tasks.  Tasks that are in the process of being
1640 <     * submitted or executed concurrently during the course of this
1641 <     * method may or may not be rejected. Unlike some other executors,
1642 <     * this method cancels rather than collects non-executed tasks
1643 <     * upon termination, so always returns an empty list. However, you
1644 <     * can use method {@code drainTasksTo} before invoking this
1645 <     * method to transfer unexecuted tasks to another collection.
1638 >     * Attempts to cancel and/or stop all tasks, and reject all
1639 >     * subsequently submitted tasks.  Tasks that are in the process of
1640 >     * being submitted or executed concurrently during the course of
1641 >     * this method may or may not be rejected. This method cancels
1642 >     * both existing and unexecuted tasks, in order to permit
1643 >     * termination in the presence of task dependencies. So the method
1644 >     * always returns an empty list (unlike the case for some other
1645 >     * Executors).
1646 >     *
1647       * @return an empty list
1648       * @throws SecurityException if a security manager exists and
1649       *         the caller is not permitted to modify threads
1650       *         because it does not hold {@link
1651 <     *         java.lang.RuntimePermission}{@code ("modifyThread")},
1651 >     *         java.lang.RuntimePermission}{@code ("modifyThread")}
1652       */
1653      public List<Runnable> shutdownNow() {
1654          checkPermission();
1655 <        terminate();
1655 >        tryTerminate(true);
1656          return Collections.emptyList();
1657      }
1658  
# Line 1085 | Line 1662 | public class ForkJoinPool extends Abstra
1662       * @return {@code true} if all tasks have completed following shut down
1663       */
1664      public boolean isTerminated() {
1665 <        return runStateOf(runControl) == TERMINATED;
1665 >        return runState >= TERMINATED;
1666      }
1667  
1668      /**
1669       * Returns {@code true} if the process of termination has
1670 <     * commenced but possibly not yet completed.
1670 >     * commenced but not yet completed.  This method may be useful for
1671 >     * debugging. A return of {@code true} reported a sufficient
1672 >     * period after shutdown may indicate that submitted tasks have
1673 >     * ignored or suppressed interruption, causing this executor not
1674 >     * to properly terminate.
1675       *
1676 <     * @return {@code true} if terminating
1676 >     * @return {@code true} if terminating but not yet terminated
1677       */
1678      public boolean isTerminating() {
1679 <        return runStateOf(runControl) >= TERMINATING;
1679 >        return (runState & (TERMINATING|TERMINATED)) == TERMINATING;
1680      }
1681  
1682      /**
# Line 1104 | Line 1685 | public class ForkJoinPool extends Abstra
1685       * @return {@code true} if this pool has been shut down
1686       */
1687      public boolean isShutdown() {
1688 <        return runStateOf(runControl) >= SHUTDOWN;
1688 >        return runState >= SHUTDOWN;
1689      }
1690  
1691      /**
# Line 1120 | Line 1701 | public class ForkJoinPool extends Abstra
1701       */
1702      public boolean awaitTermination(long timeout, TimeUnit unit)
1703          throws InterruptedException {
1123        long nanos = unit.toNanos(timeout);
1124        final ReentrantLock lock = this.workerLock;
1125        lock.lock();
1704          try {
1705 <            for (;;) {
1706 <                if (isTerminated())
1129 <                    return true;
1130 <                if (nanos <= 0)
1131 <                    return false;
1132 <                nanos = termination.awaitNanos(nanos);
1133 <            }
1134 <        } finally {
1135 <            lock.unlock();
1136 <        }
1137 <    }
1138 <
1139 <    // Shutdown and termination support
1140 <
1141 <    /**
1142 <     * Callback from terminating worker. Null out the corresponding
1143 <     * workers slot, and if terminating, try to terminate, else try to
1144 <     * shrink workers array.
1145 <     * @param w the worker
1146 <     */
1147 <    final void workerTerminated(ForkJoinWorkerThread w) {
1148 <        updateStealCount(w);
1149 <        updateWorkerCount(-1);
1150 <        final ReentrantLock lock = this.workerLock;
1151 <        lock.lock();
1152 <        try {
1153 <            ForkJoinWorkerThread[] ws = workers;
1154 <            if (ws != null) {
1155 <                int idx = w.poolIndex;
1156 <                if (idx >= 0 && idx < ws.length && ws[idx] == w)
1157 <                    ws[idx] = null;
1158 <                if (totalCountOf(workerCounts) == 0) {
1159 <                    terminate(); // no-op if already terminating
1160 <                    transitionRunStateTo(TERMINATED);
1161 <                    termination.signalAll();
1162 <                }
1163 <                else if (!isTerminating()) {
1164 <                    tryShrinkWorkerArray();
1165 <                    tryResumeSpare(true); // allow replacement
1166 <                }
1167 <            }
1168 <        } finally {
1169 <            lock.unlock();
1170 <        }
1171 <        signalIdleWorkers();
1172 <    }
1173 <
1174 <    /**
1175 <     * Initiate termination.
1176 <     */
1177 <    private void terminate() {
1178 <        if (transitionRunStateTo(TERMINATING)) {
1179 <            stopAllWorkers();
1180 <            resumeAllSpares();
1181 <            signalIdleWorkers();
1182 <            cancelQueuedSubmissions();
1183 <            cancelQueuedWorkerTasks();
1184 <            interruptUnterminatedWorkers();
1185 <            signalIdleWorkers(); // resignal after interrupt
1186 <        }
1187 <    }
1188 <
1189 <    /**
1190 <     * Possibly terminates when on shutdown state.
1191 <     */
1192 <    private void terminateOnShutdown() {
1193 <        if (!hasQueuedSubmissions() && canTerminateOnShutdown(runControl))
1194 <            terminate();
1195 <    }
1196 <
1197 <    /**
1198 <     * Clears out and cancels submissions.
1199 <     */
1200 <    private void cancelQueuedSubmissions() {
1201 <        ForkJoinTask<?> task;
1202 <        while ((task = pollSubmission()) != null)
1203 <            task.cancel(false);
1204 <    }
1205 <
1206 <    /**
1207 <     * Cleans out worker queues.
1208 <     */
1209 <    private void cancelQueuedWorkerTasks() {
1210 <        final ReentrantLock lock = this.workerLock;
1211 <        lock.lock();
1212 <        try {
1213 <            ForkJoinWorkerThread[] ws = workers;
1214 <            if (ws != null) {
1215 <                for (int i = 0; i < ws.length; ++i) {
1216 <                    ForkJoinWorkerThread t = ws[i];
1217 <                    if (t != null)
1218 <                        t.cancelTasks();
1219 <                }
1220 <            }
1221 <        } finally {
1222 <            lock.unlock();
1223 <        }
1224 <    }
1225 <
1226 <    /**
1227 <     * Sets each worker's status to terminating. Requires lock to avoid
1228 <     * conflicts with add/remove.
1229 <     */
1230 <    private void stopAllWorkers() {
1231 <        final ReentrantLock lock = this.workerLock;
1232 <        lock.lock();
1233 <        try {
1234 <            ForkJoinWorkerThread[] ws = workers;
1235 <            if (ws != null) {
1236 <                for (int i = 0; i < ws.length; ++i) {
1237 <                    ForkJoinWorkerThread t = ws[i];
1238 <                    if (t != null)
1239 <                        t.shutdownNow();
1240 <                }
1241 <            }
1242 <        } finally {
1243 <            lock.unlock();
1244 <        }
1245 <    }
1246 <
1247 <    /**
1248 <     * Interrupts all unterminated workers.  This is not required for
1249 <     * sake of internal control, but may help unstick user code during
1250 <     * shutdown.
1251 <     */
1252 <    private void interruptUnterminatedWorkers() {
1253 <        final ReentrantLock lock = this.workerLock;
1254 <        lock.lock();
1255 <        try {
1256 <            ForkJoinWorkerThread[] ws = workers;
1257 <            if (ws != null) {
1258 <                for (int i = 0; i < ws.length; ++i) {
1259 <                    ForkJoinWorkerThread t = ws[i];
1260 <                    if (t != null && !t.isTerminated()) {
1261 <                        try {
1262 <                            t.interrupt();
1263 <                        } catch (SecurityException ignore) {
1264 <                        }
1265 <                    }
1266 <                }
1267 <            }
1268 <        } finally {
1269 <            lock.unlock();
1270 <        }
1271 <    }
1272 <
1273 <
1274 <    /*
1275 <     * Nodes for event barrier to manage idle threads.  Queue nodes
1276 <     * are basic Treiber stack nodes, also used for spare stack.
1277 <     *
1278 <     * The event barrier has an event count and a wait queue (actually
1279 <     * a Treiber stack).  Workers are enabled to look for work when
1280 <     * the eventCount is incremented. If they fail to find work, they
1281 <     * may wait for next count. Upon release, threads help others wake
1282 <     * up.
1283 <     *
1284 <     * Synchronization events occur only in enough contexts to
1285 <     * maintain overall liveness:
1286 <     *
1287 <     *   - Submission of a new task to the pool
1288 <     *   - Resizes or other changes to the workers array
1289 <     *   - pool termination
1290 <     *   - A worker pushing a task on an empty queue
1291 <     *
1292 <     * The case of pushing a task occurs often enough, and is heavy
1293 <     * enough compared to simple stack pushes, to require special
1294 <     * handling: Method signalWork returns without advancing count if
1295 <     * the queue appears to be empty.  This would ordinarily result in
1296 <     * races causing some queued waiters not to be woken up. To avoid
1297 <     * this, the first worker enqueued in method sync (see
1298 <     * syncIsReleasable) rescans for tasks after being enqueued, and
1299 <     * helps signal if any are found. This works well because the
1300 <     * worker has nothing better to do, and so might as well help
1301 <     * alleviate the overhead and contention on the threads actually
1302 <     * doing work.  Also, since event counts increments on task
1303 <     * availability exist to maintain liveness (rather than to force
1304 <     * refreshes etc), it is OK for callers to exit early if
1305 <     * contending with another signaller.
1306 <     */
1307 <    static final class WaitQueueNode {
1308 <        WaitQueueNode next; // only written before enqueued
1309 <        volatile ForkJoinWorkerThread thread; // nulled to cancel wait
1310 <        final long count; // unused for spare stack
1311 <
1312 <        WaitQueueNode(long c, ForkJoinWorkerThread w) {
1313 <            count = c;
1314 <            thread = w;
1315 <        }
1316 <
1317 <        /**
1318 <         * Wakes up waiter, returning false if known to already
1319 <         */
1320 <        boolean signal() {
1321 <            ForkJoinWorkerThread t = thread;
1322 <            if (t == null)
1323 <                return false;
1324 <            thread = null;
1325 <            LockSupport.unpark(t);
1326 <            return true;
1327 <        }
1328 <
1329 <        /**
1330 <         * Awaits release on sync.
1331 <         */
1332 <        void awaitSyncRelease(ForkJoinPool p) {
1333 <            while (thread != null && !p.syncIsReleasable(this))
1334 <                LockSupport.park(this);
1335 <        }
1336 <
1337 <        /**
1338 <         * Awaits resumption as spare.
1339 <         */
1340 <        void awaitSpareRelease() {
1341 <            while (thread != null) {
1342 <                if (!Thread.interrupted())
1343 <                    LockSupport.park(this);
1344 <            }
1345 <        }
1346 <    }
1347 <
1348 <    /**
1349 <     * Ensures that no thread is waiting for count to advance from the
1350 <     * current value of eventCount read on entry to this method, by
1351 <     * releasing waiting threads if necessary.
1352 <     * @return the count
1353 <     */
1354 <    final long ensureSync() {
1355 <        long c = eventCount;
1356 <        WaitQueueNode q;
1357 <        while ((q = syncStack) != null && q.count < c) {
1358 <            if (casBarrierStack(q, null)) {
1359 <                do {
1360 <                    q.signal();
1361 <                } while ((q = q.next) != null);
1362 <                break;
1363 <            }
1364 <        }
1365 <        return c;
1366 <    }
1367 <
1368 <    /**
1369 <     * Increments event count and releases waiting threads.
1370 <     */
1371 <    private void signalIdleWorkers() {
1372 <        long c;
1373 <        do;while (!casEventCount(c = eventCount, c+1));
1374 <        ensureSync();
1375 <    }
1376 <
1377 <    /**
1378 <     * Signals threads waiting to poll a task. Because method sync
1379 <     * rechecks availability, it is OK to only proceed if queue
1380 <     * appears to be non-empty, and OK to skip under contention to
1381 <     * increment count (since some other thread succeeded).
1382 <     */
1383 <    final void signalWork() {
1384 <        long c;
1385 <        WaitQueueNode q;
1386 <        if (syncStack != null &&
1387 <            casEventCount(c = eventCount, c+1) &&
1388 <            (((q = syncStack) != null && q.count <= c) &&
1389 <             (!casBarrierStack(q, q.next) || !q.signal())))
1390 <            ensureSync();
1391 <    }
1392 <
1393 <    /**
1394 <     * Waits until event count advances from last value held by
1395 <     * caller, or if excess threads, caller is resumed as spare, or
1396 <     * caller or pool is terminating. Updates caller's event on exit.
1397 <     * @param w the calling worker thread
1398 <     */
1399 <    final void sync(ForkJoinWorkerThread w) {
1400 <        updateStealCount(w); // Transfer w's count while it is idle
1401 <
1402 <        while (!w.isShutdown() && !isTerminating() && !suspendIfSpare(w)) {
1403 <            long prev = w.lastEventCount;
1404 <            WaitQueueNode node = null;
1405 <            WaitQueueNode h;
1406 <            while (eventCount == prev &&
1407 <                   ((h = syncStack) == null || h.count == prev)) {
1408 <                if (node == null)
1409 <                    node = new WaitQueueNode(prev, w);
1410 <                if (casBarrierStack(node.next = h, node)) {
1411 <                    node.awaitSyncRelease(this);
1412 <                    break;
1413 <                }
1414 <            }
1415 <            long ec = ensureSync();
1416 <            if (ec != prev) {
1417 <                w.lastEventCount = ec;
1418 <                break;
1419 <            }
1420 <        }
1421 <    }
1422 <
1423 <    /**
1424 <     * Returns true if worker waiting on sync can proceed:
1425 <     *  - on signal (thread == null)
1426 <     *  - on event count advance (winning race to notify vs signaller)
1427 <     *  - on Interrupt
1428 <     *  - if the first queued node, we find work available
1429 <     * If node was not signalled and event count not advanced on exit,
1430 <     * then we also help advance event count.
1431 <     * @return true if node can be released
1432 <     */
1433 <    final boolean syncIsReleasable(WaitQueueNode node) {
1434 <        long prev = node.count;
1435 <        if (!Thread.interrupted() && node.thread != null &&
1436 <            (node.next != null ||
1437 <             !ForkJoinWorkerThread.hasQueuedTasks(workers)) &&
1438 <            eventCount == prev)
1439 <            return false;
1440 <        if (node.thread != null) {
1441 <            node.thread = null;
1442 <            long ec = eventCount;
1443 <            if (prev <= ec) // help signal
1444 <                casEventCount(ec, ec+1);
1445 <        }
1446 <        return true;
1447 <    }
1448 <
1449 <    /**
1450 <     * Returns true if a new sync event occurred since last call to
1451 <     * sync or this method, if so, updating caller's count.
1452 <     */
1453 <    final boolean hasNewSyncEvent(ForkJoinWorkerThread w) {
1454 <        long lc = w.lastEventCount;
1455 <        long ec = ensureSync();
1456 <        if (ec == lc)
1705 >            return termination.awaitAdvanceInterruptibly(0, timeout, unit) > 0;
1706 >        } catch(TimeoutException ex) {
1707              return false;
1458        w.lastEventCount = ec;
1459        return true;
1460    }
1461
1462    //  Parallelism maintenance
1463
1464    /**
1465     * Decrements running count; if too low, adds spare.
1466     *
1467     * Conceptually, all we need to do here is add or resume a
1468     * spare thread when one is about to block (and remove or
1469     * suspend it later when unblocked -- see suspendIfSpare).
1470     * However, implementing this idea requires coping with
1471     * several problems: We have imperfect information about the
1472     * states of threads. Some count updates can and usually do
1473     * lag run state changes, despite arrangements to keep them
1474     * accurate (for example, when possible, updating counts
1475     * before signalling or resuming), especially when running on
1476     * dynamic JVMs that don't optimize the infrequent paths that
1477     * update counts. Generating too many threads can make these
1478     * problems become worse, because excess threads are more
1479     * likely to be context-switched with others, slowing them all
1480     * down, especially if there is no work available, so all are
1481     * busy scanning or idling.  Also, excess spare threads can
1482     * only be suspended or removed when they are idle, not
1483     * immediately when they aren't needed. So adding threads will
1484     * raise parallelism level for longer than necessary.  Also,
1485     * FJ applications often encounter highly transient peaks when
1486     * many threads are blocked joining, but for less time than it
1487     * takes to create or resume spares.
1488     *
1489     * @param joinMe if non-null, return early if done
1490     * @param maintainParallelism if true, try to stay within
1491     * target counts, else create only to avoid starvation
1492     * @return true if joinMe known to be done
1493     */
1494    final boolean preJoin(ForkJoinTask<?> joinMe, boolean maintainParallelism) {
1495        maintainParallelism &= maintainsParallelism; // overrride
1496        boolean dec = false;  // true when running count decremented
1497        while (spareStack == null || !tryResumeSpare(dec)) {
1498            int counts = workerCounts;
1499            if (dec || (dec = casWorkerCounts(counts, --counts))) { // CAS cheat
1500                if (!needSpare(counts, maintainParallelism))
1501                    break;
1502                if (joinMe.status < 0)
1503                    return true;
1504                if (tryAddSpare(counts))
1505                    break;
1506            }
1507        }
1508        return false;
1509    }
1510
1511    /**
1512     * Same idea as preJoin
1513     */
1514    final boolean preBlock(ManagedBlocker blocker, boolean maintainParallelism){
1515        maintainParallelism &= maintainsParallelism;
1516        boolean dec = false;
1517        while (spareStack == null || !tryResumeSpare(dec)) {
1518            int counts = workerCounts;
1519            if (dec || (dec = casWorkerCounts(counts, --counts))) {
1520                if (!needSpare(counts, maintainParallelism))
1521                    break;
1522                if (blocker.isReleasable())
1523                    return true;
1524                if (tryAddSpare(counts))
1525                    break;
1526            }
1527        }
1528        return false;
1529    }
1530
1531    /**
1532     * Returns true if a spare thread appears to be needed.  If
1533     * maintaining parallelism, returns true when the deficit in
1534     * running threads is more than the surplus of total threads, and
1535     * there is apparently some work to do.  This self-limiting rule
1536     * means that the more threads that have already been added, the
1537     * less parallelism we will tolerate before adding another.
1538     * @param counts current worker counts
1539     * @param maintainParallelism try to maintain parallelism
1540     */
1541    private boolean needSpare(int counts, boolean maintainParallelism) {
1542        int ps = parallelism;
1543        int rc = runningCountOf(counts);
1544        int tc = totalCountOf(counts);
1545        int runningDeficit = ps - rc;
1546        int totalSurplus = tc - ps;
1547        return (tc < maxPoolSize &&
1548                (rc == 0 || totalSurplus < 0 ||
1549                 (maintainParallelism &&
1550                  runningDeficit > totalSurplus &&
1551                  ForkJoinWorkerThread.hasQueuedTasks(workers))));
1552    }
1553
1554    /**
1555     * Adds a spare worker if lock available and no more than the
1556     * expected numbers of threads exist.
1557     * @return true if successful
1558     */
1559    private boolean tryAddSpare(int expectedCounts) {
1560        final ReentrantLock lock = this.workerLock;
1561        int expectedRunning = runningCountOf(expectedCounts);
1562        int expectedTotal = totalCountOf(expectedCounts);
1563        boolean success = false;
1564        boolean locked = false;
1565        // confirm counts while locking; CAS after obtaining lock
1566        try {
1567            for (;;) {
1568                int s = workerCounts;
1569                int tc = totalCountOf(s);
1570                int rc = runningCountOf(s);
1571                if (rc > expectedRunning || tc > expectedTotal)
1572                    break;
1573                if (!locked && !(locked = lock.tryLock()))
1574                    break;
1575                if (casWorkerCounts(s, workerCountsFor(tc+1, rc+1))) {
1576                    createAndStartSpare(tc);
1577                    success = true;
1578                    break;
1579                }
1580            }
1581        } finally {
1582            if (locked)
1583                lock.unlock();
1584        }
1585        return success;
1586    }
1587
1588    /**
1589     * Adds the kth spare worker. On entry, pool counts are already
1590     * adjusted to reflect addition.
1591     */
1592    private void createAndStartSpare(int k) {
1593        ForkJoinWorkerThread w = null;
1594        ForkJoinWorkerThread[] ws = ensureWorkerArrayCapacity(k + 1);
1595        int len = ws.length;
1596        // Probably, we can place at slot k. If not, find empty slot
1597        if (k < len && ws[k] != null) {
1598            for (k = 0; k < len && ws[k] != null; ++k)
1599                ;
1600        }
1601        if (k < len && !isTerminating() && (w = createWorker(k)) != null) {
1602            ws[k] = w;
1603            w.start();
1604        }
1605        else
1606            updateWorkerCount(-1); // adjust on failure
1607        signalIdleWorkers();
1608    }
1609
1610    /**
1611     * Suspends calling thread w if there are excess threads.  Called
1612     * only from sync.  Spares are enqueued in a Treiber stack using
1613     * the same WaitQueueNodes as barriers.  They are resumed mainly
1614     * in preJoin, but are also woken on pool events that require all
1615     * threads to check run state.
1616     * @param w the caller
1617     */
1618    private boolean suspendIfSpare(ForkJoinWorkerThread w) {
1619        WaitQueueNode node = null;
1620        int s;
1621        while (parallelism < runningCountOf(s = workerCounts)) {
1622            if (node == null)
1623                node = new WaitQueueNode(0, w);
1624            if (casWorkerCounts(s, s-1)) { // representation-dependent
1625                // push onto stack
1626                do;while (!casSpareStack(node.next = spareStack, node));
1627                // block until released by resumeSpare
1628                node.awaitSpareRelease();
1629                return true;
1630            }
1631        }
1632        return false;
1633    }
1634
1635    /**
1636     * Tries to pop and resume a spare thread.
1637     * @param updateCount if true, increment running count on success
1638     * @return true if successful
1639     */
1640    private boolean tryResumeSpare(boolean updateCount) {
1641        WaitQueueNode q;
1642        while ((q = spareStack) != null) {
1643            if (casSpareStack(q, q.next)) {
1644                if (updateCount)
1645                    updateRunningCount(1);
1646                q.signal();
1647                return true;
1648            }
1649        }
1650        return false;
1651    }
1652
1653    /**
1654     * Pops and resumes all spare threads. Same idea as ensureSync.
1655     * @return true if any spares released
1656     */
1657    private boolean resumeAllSpares() {
1658        WaitQueueNode q;
1659        while ( (q = spareStack) != null) {
1660            if (casSpareStack(q, null)) {
1661                do {
1662                    updateRunningCount(1);
1663                    q.signal();
1664                } while ((q = q.next) != null);
1665                return true;
1666            }
1667        }
1668        return false;
1669    }
1670
1671    /**
1672     * Pops and shuts down excessive spare threads. Call only while
1673     * holding lock. This is not guaranteed to eliminate all excess
1674     * threads, only those suspended as spares, which are the ones
1675     * unlikely to be needed in the future.
1676     */
1677    private void trimSpares() {
1678        int surplus = totalCountOf(workerCounts) - parallelism;
1679        WaitQueueNode q;
1680        while (surplus > 0 && (q = spareStack) != null) {
1681            if (casSpareStack(q, null)) {
1682                do {
1683                    updateRunningCount(1);
1684                    ForkJoinWorkerThread w = q.thread;
1685                    if (w != null && surplus > 0 &&
1686                        runningCountOf(workerCounts) > 0 && w.shutdown())
1687                        --surplus;
1688                    q.signal();
1689                } while ((q = q.next) != null);
1690            }
1708          }
1709      }
1710  
1711      /**
1712       * Interface for extending managed parallelism for tasks running
1713 <     * in ForkJoinPools. A ManagedBlocker provides two methods.
1714 <     * Method {@code isReleasable} must return true if blocking is not
1715 <     * necessary. Method {@code block} blocks the current thread
1716 <     * if necessary (perhaps internally invoking isReleasable before
1717 <     * actually blocking.).
1713 >     * in {@link ForkJoinPool}s.
1714 >     *
1715 >     * <p>A {@code ManagedBlocker} provides two methods.  Method
1716 >     * {@code isReleasable} must return {@code true} if blocking is
1717 >     * not necessary. Method {@code block} blocks the current thread
1718 >     * if necessary (perhaps internally invoking {@code isReleasable}
1719 >     * before actually blocking). The unusual methods in this API
1720 >     * accommodate synchronizers that may, but don't usually, block
1721 >     * for long periods. Similarly, they allow more efficient internal
1722 >     * handling of cases in which additional workers may be, but
1723 >     * usually are not, needed to ensure sufficient parallelism.
1724 >     * Toward this end, implementations of method {@code isReleasable}
1725 >     * must be amenable to repeated invocation.
1726 >     *
1727       * <p>For example, here is a ManagedBlocker based on a
1728       * ReentrantLock:
1729 <     * <pre>
1730 <     *   class ManagedLocker implements ManagedBlocker {
1731 <     *     final ReentrantLock lock;
1732 <     *     boolean hasLock = false;
1733 <     *     ManagedLocker(ReentrantLock lock) { this.lock = lock; }
1734 <     *     public boolean block() {
1735 <     *        if (!hasLock)
1736 <     *           lock.lock();
1737 <     *        return true;
1738 <     *     }
1739 <     *     public boolean isReleasable() {
1740 <     *        return hasLock || (hasLock = lock.tryLock());
1741 <     *     }
1729 >     *  <pre> {@code
1730 >     * class ManagedLocker implements ManagedBlocker {
1731 >     *   final ReentrantLock lock;
1732 >     *   boolean hasLock = false;
1733 >     *   ManagedLocker(ReentrantLock lock) { this.lock = lock; }
1734 >     *   public boolean block() {
1735 >     *     if (!hasLock)
1736 >     *       lock.lock();
1737 >     *     return true;
1738 >     *   }
1739 >     *   public boolean isReleasable() {
1740 >     *     return hasLock || (hasLock = lock.tryLock());
1741 >     *   }
1742 >     * }}</pre>
1743 >     *
1744 >     * <p>Here is a class that possibly blocks waiting for an
1745 >     * item on a given queue:
1746 >     *  <pre> {@code
1747 >     * class QueueTaker<E> implements ManagedBlocker {
1748 >     *   final BlockingQueue<E> queue;
1749 >     *   volatile E item = null;
1750 >     *   QueueTaker(BlockingQueue<E> q) { this.queue = q; }
1751 >     *   public boolean block() throws InterruptedException {
1752 >     *     if (item == null)
1753 >     *       item = queue.take
1754 >     *     return true;
1755       *   }
1756 <     * </pre>
1756 >     *   public boolean isReleasable() {
1757 >     *     return item != null || (item = queue.poll) != null;
1758 >     *   }
1759 >     *   public E getItem() { // call after pool.managedBlock completes
1760 >     *     return item;
1761 >     *   }
1762 >     * }}</pre>
1763       */
1764      public static interface ManagedBlocker {
1765          /**
1766           * Possibly blocks the current thread, for example waiting for
1767           * a lock or condition.
1768 <         * @return true if no additional blocking is necessary (i.e.,
1769 <         * if isReleasable would return true)
1768 >         *
1769 >         * @return {@code true} if no additional blocking is necessary
1770 >         * (i.e., if isReleasable would return true)
1771           * @throws InterruptedException if interrupted while waiting
1772 <         * (the method is not required to do so, but is allowed to).
1772 >         * (the method is not required to do so, but is allowed to)
1773           */
1774          boolean block() throws InterruptedException;
1775  
1776          /**
1777 <         * Returns true if blocking is unnecessary.
1777 >         * Returns {@code true} if blocking is unnecessary.
1778           */
1779          boolean isReleasable();
1780      }
1781  
1782      /**
1783       * Blocks in accord with the given blocker.  If the current thread
1784 <     * is a ForkJoinWorkerThread, this method possibly arranges for a
1785 <     * spare thread to be activated if necessary to ensure parallelism
1786 <     * while the current thread is blocked.  If
1787 <     * {@code maintainParallelism} is true and the pool supports
1788 <     * it ({@link #getMaintainsParallelism}), this method attempts to
1789 <     * maintain the pool's nominal parallelism. Otherwise if activates
1790 <     * a thread only if necessary to avoid complete starvation. This
1791 <     * option may be preferable when blockages use timeouts, or are
1792 <     * almost always brief.
1793 <     *
1794 <     * <p> If the caller is not a ForkJoinTask, this method is behaviorally
1795 <     * equivalent to
1796 <     * <pre>
1797 <     *   while (!blocker.isReleasable())
1752 <     *      if (blocker.block())
1753 <     *         return;
1754 <     * </pre>
1755 <     * If the caller is a ForkJoinTask, then the pool may first
1756 <     * be expanded to ensure parallelism, and later adjusted.
1784 >     * is a {@link ForkJoinWorkerThread}, this method possibly
1785 >     * arranges for a spare thread to be activated if necessary to
1786 >     * ensure sufficient parallelism while the current thread is blocked.
1787 >     *
1788 >     * <p>If the caller is not a {@link ForkJoinTask}, this method is
1789 >     * behaviorally equivalent to
1790 >     *  <pre> {@code
1791 >     * while (!blocker.isReleasable())
1792 >     *   if (blocker.block())
1793 >     *     return;
1794 >     * }</pre>
1795 >     *
1796 >     * If the caller is a {@code ForkJoinTask}, then the pool may
1797 >     * first be expanded to ensure parallelism, and later adjusted.
1798       *
1799       * @param blocker the blocker
1759     * @param maintainParallelism if true and supported by this pool,
1760     * attempt to maintain the pool's nominal parallelism; otherwise
1761     * activate a thread only if necessary to avoid complete
1762     * starvation.
1800       * @throws InterruptedException if blocker.block did so
1801       */
1802 <    public static void managedBlock(ManagedBlocker blocker,
1766 <                                    boolean maintainParallelism)
1802 >    public static void managedBlock(ManagedBlocker blocker)
1803          throws InterruptedException {
1804          Thread t = Thread.currentThread();
1805 <        ForkJoinPool pool = (t instanceof ForkJoinWorkerThread?
1806 <                             ((ForkJoinWorkerThread)t).pool : null);
1807 <        if (!blocker.isReleasable()) {
1808 <            try {
1809 <                if (pool == null ||
1810 <                    !pool.preBlock(blocker, maintainParallelism))
1775 <                    awaitBlocker(blocker);
1776 <            } finally {
1777 <                if (pool != null)
1778 <                    pool.updateRunningCount(1);
1779 <            }
1805 >        if (t instanceof ForkJoinWorkerThread) {
1806 >            ForkJoinWorkerThread w = (ForkJoinWorkerThread) t;
1807 >            w.pool.awaitBlocker(blocker);
1808 >        }
1809 >        else {
1810 >            do {} while (!blocker.isReleasable() && !blocker.block());
1811          }
1812      }
1813  
1814 <    private static void awaitBlocker(ManagedBlocker blocker)
1815 <        throws InterruptedException {
1816 <        do;while (!blocker.isReleasable() && !blocker.block());
1786 <    }
1787 <
1788 <    // AbstractExecutorService overrides
1814 >    // AbstractExecutorService overrides.  These rely on undocumented
1815 >    // fact that ForkJoinTask.adapt returns ForkJoinTasks that also
1816 >    // implement RunnableFuture.
1817  
1818      protected <T> RunnableFuture<T> newTaskFor(Runnable runnable, T value) {
1819 <        return new AdaptedRunnable(runnable, value);
1819 >        return (RunnableFuture<T>) ForkJoinTask.adapt(runnable, value);
1820      }
1821  
1822      protected <T> RunnableFuture<T> newTaskFor(Callable<T> callable) {
1823 <        return new AdaptedCallable(callable);
1823 >        return (RunnableFuture<T>) ForkJoinTask.adapt(callable);
1824      }
1825  
1826 +    // Unsafe mechanics
1827  
1828 <    // Temporary Unsafe mechanics for preliminary release
1829 <    private static Unsafe getUnsafe() throws Throwable {
1828 >    private static final sun.misc.Unsafe UNSAFE = getUnsafe();
1829 >    private static final long workerCountsOffset =
1830 >        objectFieldOffset("workerCounts", ForkJoinPool.class);
1831 >    private static final long runStateOffset =
1832 >        objectFieldOffset("runState", ForkJoinPool.class);
1833 >    private static final long eventCountOffset =
1834 >        objectFieldOffset("eventCount", ForkJoinPool.class);
1835 >    private static final long eventWaitersOffset =
1836 >        objectFieldOffset("eventWaiters",ForkJoinPool.class);
1837 >    private static final long stealCountOffset =
1838 >        objectFieldOffset("stealCount",ForkJoinPool.class);
1839 >    private static final long spareWaitersOffset =
1840 >        objectFieldOffset("spareWaiters",ForkJoinPool.class);
1841 >
1842 >    private static long objectFieldOffset(String field, Class<?> klazz) {
1843 >        try {
1844 >            return UNSAFE.objectFieldOffset(klazz.getDeclaredField(field));
1845 >        } catch (NoSuchFieldException e) {
1846 >            // Convert Exception to corresponding Error
1847 >            NoSuchFieldError error = new NoSuchFieldError(field);
1848 >            error.initCause(e);
1849 >            throw error;
1850 >        }
1851 >    }
1852 >
1853 >    /**
1854 >     * Returns a sun.misc.Unsafe.  Suitable for use in a 3rd party package.
1855 >     * Replace with a simple call to Unsafe.getUnsafe when integrating
1856 >     * into a jdk.
1857 >     *
1858 >     * @return a sun.misc.Unsafe
1859 >     */
1860 >    private static sun.misc.Unsafe getUnsafe() {
1861          try {
1862 <            return Unsafe.getUnsafe();
1862 >            return sun.misc.Unsafe.getUnsafe();
1863          } catch (SecurityException se) {
1864              try {
1865                  return java.security.AccessController.doPrivileged
1866 <                    (new java.security.PrivilegedExceptionAction<Unsafe>() {
1867 <                        public Unsafe run() throws Exception {
1868 <                            return getUnsafePrivileged();
1866 >                    (new java.security
1867 >                     .PrivilegedExceptionAction<sun.misc.Unsafe>() {
1868 >                        public sun.misc.Unsafe run() throws Exception {
1869 >                            java.lang.reflect.Field f = sun.misc
1870 >                                .Unsafe.class.getDeclaredField("theUnsafe");
1871 >                            f.setAccessible(true);
1872 >                            return (sun.misc.Unsafe) f.get(null);
1873                          }});
1874              } catch (java.security.PrivilegedActionException e) {
1875 <                throw e.getCause();
1875 >                throw new RuntimeException("Could not initialize intrinsics",
1876 >                                           e.getCause());
1877              }
1878          }
1879      }
1815
1816    private static Unsafe getUnsafePrivileged()
1817            throws NoSuchFieldException, IllegalAccessException {
1818        Field f = Unsafe.class.getDeclaredField("theUnsafe");
1819        f.setAccessible(true);
1820        return (Unsafe) f.get(null);
1821    }
1822
1823    private static long fieldOffset(String fieldName)
1824            throws NoSuchFieldException {
1825        return UNSAFE.objectFieldOffset
1826            (ForkJoinPool.class.getDeclaredField(fieldName));
1827    }
1828
1829    static final Unsafe UNSAFE;
1830    static final long eventCountOffset;
1831    static final long workerCountsOffset;
1832    static final long runControlOffset;
1833    static final long syncStackOffset;
1834    static final long spareStackOffset;
1835
1836    static {
1837        try {
1838            UNSAFE = getUnsafe();
1839            eventCountOffset = fieldOffset("eventCount");
1840            workerCountsOffset = fieldOffset("workerCounts");
1841            runControlOffset = fieldOffset("runControl");
1842            syncStackOffset = fieldOffset("syncStack");
1843            spareStackOffset = fieldOffset("spareStack");
1844        } catch (Throwable e) {
1845            throw new RuntimeException("Could not initialize intrinsics", e);
1846        }
1847    }
1848
1849    private boolean casEventCount(long cmp, long val) {
1850        return UNSAFE.compareAndSwapLong(this, eventCountOffset, cmp, val);
1851    }
1852    private boolean casWorkerCounts(int cmp, int val) {
1853        return UNSAFE.compareAndSwapInt(this, workerCountsOffset, cmp, val);
1854    }
1855    private boolean casRunControl(int cmp, int val) {
1856        return UNSAFE.compareAndSwapInt(this, runControlOffset, cmp, val);
1857    }
1858    private boolean casSpareStack(WaitQueueNode cmp, WaitQueueNode val) {
1859        return UNSAFE.compareAndSwapObject(this, spareStackOffset, cmp, val);
1860    }
1861    private boolean casBarrierStack(WaitQueueNode cmp, WaitQueueNode val) {
1862        return UNSAFE.compareAndSwapObject(this, syncStackOffset, cmp, val);
1863    }
1880   }

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines